Abstract
As more patients with established rheumatoid arthritis (RA) achieve remission or low disease activity, strategies such as tapering and withdrawal of disease-modifying antirheumatic drugs (DMARDs) are being investigated. In several trials, DMARD discontinuation was associated with a higher risk of relapse, ranging from 56% to 87% at 1 year. Tapering, either by dose reduction or by injection spacing, may limit the risk of relapse. Half-dose etanercept (ETN) versus full-dose continuation was not associated with an increased relapse risk at 1 year in two trials. Progressive antitumor necrosis factor injection spacing was shown to be equivalent to full regimen continuation in terms of persistent flare and disease activity at 18 months in one trial, but not in another one. Reintroduction of a DMARD at previous dose/regimen was usually associated with remission re-induction. The risk of relevant structural damage progression was not increased. Safety improvement has not yet been demonstrated. The annual cost reduction when tapering biologic DMARDs (bDMARDs) was 3500–6000 €/patient. Research questions to be addressed include defining flare that requires reinitiation of treatment, such that patients facilitate the maintenance of remission during tapering by timely communication with their rheumatology team.
Outcomes for established rheumatoid arthritis (RA) patients have substantially improved during the recent years. This has been accomplished thanks to numerous drug developments – new conventional synthetic and biologic disease-modifying antirheumatic drugs (csDMARDs and bDMARD, respectively) – as well as by implementation of early aggressive and dynamic treatments, that is, step-up and treat-to-target strategies based on therapy adjustment until low disease activity (LDA) or remission is achieved. This has made remission, or at least LDA, an achievable and consensual objective for RA patients .
Achievement of LDA or sustained remission raises the possibility of DMARD tapering, titration, or/and discontinuation. Until quite recently, RA treatment was based on the paradigm of continuous treatment for a chronic and lifelong disease, with the underlying perception that drug-free remission is rare, and there is no rationale to change a winning “team” . In addition, several publications have highlighted the risk of disease relapse in cases of DMARD cessation , with the risk of subsequent secondary failure of the DMARD that was stopped. This conservative strategy contrasted with those used for other immune-mediated inflammatory diseases such as vasculitis or systemic lupus erythematosus, for which the therapeutic strategy has two phases: an induction phase with powerful immunosuppressive or -modulating agents, followed by a maintenance phase with less intensive regimens .
In the last decade, this point of view has changed for RA. The first reason was the achievement of remission or LDA in increasing numbers of patients , raising the question of whether drug-free remission would be possible to achieve . In addition, with RA prognosis improving, it appeared that the risks or disadvantages of DMARD continuation, for example, the risk of serious infections and costs, could exceed the risks of active RA. Indeed, the tapering of RA therapies has recently been included in the European League Against Rheumatism (EULAR) 2013 guidelines for the management of RA . In this review, we discuss why, how, and in whom treatment relaxation can be attempted, summarizing the current base of evidence.
The “why”: reasons leading to attempt tapering of RA therapies
There are several reasons for changing the paradigm of long-term full-dose treatment in RA. First, remission is achievable in a large proportion of patients in daily practice – for example, up to 47% of RA patients for Disease Activity Score – 28 joints (DAS28) achieved remission, reported in a recent publication based on the NOR-DMARD registry . Therefore, some rheumatologists have suggested that the maintenance of DMARDs (at full doses) for some patients could potentially represent “overtreatment,” and they advised attempting treatment reduction . Second, although remission is not presented to patients as a cure of the disease, patients who are asymptomatic on treatment may ask their physician whether “treatment holidays” or at least dose reduction could be considered at some point. This attitude reflects the human burden of long-term treatment for RA patients, which often leads to reduced adherence over time, especially when patients do not feel the signs and symptoms of the disease in their daily life activities . Third, data from both clinical trials and observational registries have clearly identified the (dose-dependent) risk of serious infections with csDMARD as well as with bDMARDs . Although this risk seems to be maximal in the few months of drug initiation, the more at-risk patients quit the treatment later. This observation is called “dilution of the susceptible” or “healthy survivor bias.” Nevertheless, this risk exists for all RA patients, and it could counterbalance the benefits of long-term therapies. Thus, dose reduction could be implemented to improve the benefit–risk profile of RA treatments.
Fourth, there could be a pharmacodynamic rationale for DMARD reduction. Most DMARDs, especially bDMARDs, work by achieving a minimal effective serum concentration . In every patient, therefore, a dose–response curve can be conceived with an LDA at a higher concentration or dose. Several patterns exist ( Fig. 1 ) as follows: (1) normal dose–response curve, (2) partial dose–response curve, (3) dose–response with curve shifted to the right (higher dose needed for effect), (4) dose–response with curve shifted to the left (lower dose is already effective), and (5) no dose–response relation (patient is doing well or not, independent of concentration/dose). Keeping these different dose–response curves in mind, in case of tapering, patients with no apparent dose–response can be tapered up to discontinuation with no expected consequences on disease activity. Patients with a dose–response curve can be tapered until the lowest effective dose, under which they reach the vertical slope and loss clinical response leading to disease activity worsening. For patients with a curve shifted to the left, this lowest effective dose is lower than for patients with a normal dose–response curve. Patients with partial response should be switched to another drug, or – when disease control is judged to be adequate – treated like patients with a normal dose–response. Of note, patients with a curve shifted to the right could receive higher than standard dosing. Based on randomized dose escalation studies, this is, however rare, and increasing the dose above registration dose also is associated with more adverse effects and unacceptable costs . Within a patient, it can be conceived that these dose–response curves remain the same over time, although sometimes, of course, secondary ineffectiveness will occur. This leads to the conclusion that patients who can eventually be stopped either never had an effect of the drug (had a spontaneous improvement of disease activity) or only had temporary need for the drug.
A potential issue when tapering is to consider that a cognitive behavioral component is introduced in the form of nocebo and attribution. When a drug is tapered, patients and physicians alike may think that a reduced dose is potentially inferior treatment, resulting in the perceived deterioration of disease control. A somewhat related effect is that of attribution. When a flare occurs after tapering, both physician and patient will most likely attribute this to the lower dose, although it could only be the natural disease evolution.
Importantly, there is an economic rationale for tapering DMARDs, especially bDMARDs: limitation of “overtreatment” using drug holidays or dose reduction may induce substantial cost savings for society, estimated at up to 1 billion euros for infliximab (IFX) alone and up to 20 billion dollars for adalimumab (ADA) and ETN .
Thus, it is possible that, in the bDMARD-treated RA population, some patients would be just as well with lower or no bDMARD, at least for a reasonable period of time. Tapering, when optimally done, would then be beneficial for both the patients and the society.
The “how”: possible strategies for tapering of RA therapies
Recent evidence for DMARD discontinuation in RA
DMARD discontinuation has been regularly tested in established RA patients in both observational studies and randomized controlled trials (RCTs) ( Table 1 ). However, substantial methodological heterogeneity was noted: enrolled patients had stable disease, defined only qualitatively in the oldest publications, or based on DAS28 LDA/remission in the most recent ones. In the latter publications, patients with DAS28 remission or LDA were selected, with a mean remission or an LDA duration of usually at least 6 months. The definition of flares or relapses differed across studies: recurrence of one or more joints with clinical synovitis, loss of DAS remission or LDA, loss of remission or LDA associated with DAS28 change of >0.6, or 1.2. Despite this heterogeneity, several observations are possible. First, observational studies were mainly conducted in patients treated with biologics in routine or in long-term extension (LTE) following industry-sponsored RCTs . The relapse rate at 1 year after bDMARD discontinuation ranged from 45% to 88%, whatever the bDMARD considered ( Table 1 ). Second, whatever the DMARD, methotrexate (MTX) or bDMARD, the risk of disease-activity exacerbation in the year following DMARD discontinuation was high, >50% except in two studies . In RCTs, comparison with the DMARD continuation arm clearly identified an excess risk of relapse, with risk ratios of ≥2 ( Table 1 ). Of note, this risk did not differ between MTX and bDMARDs, or between bDMARDs themselves. The reinitiation of the DMARD – mainly studied in bDMARD studies – was associated with an adequate new response in most patients who relapsed . Interestingly, the relapse rate when the DMARD was continued at the same dose was sometimes high, reaching 50% in two studies . One might interpret these results in several ways: (1) a residual and unavoidable risk of RA flare whatever the remission duration as part of RA natural history, which is incorrectly attributed to the tapering attempt; (2) the persistence of spontaneously resolving disease-activity fluctuations – thus not requiring DMARD adaptation or change – as was shown in previous cohort studies ; (3) the lack of reliable flare or relapse definition enabling the discrimination between the first two points ; and (4) a possible “nocebo” effect, making flare more expected when the patient is told that treatment is going to be reduced (though not in placebo-controlled studies, however).
| Reference | Drug | Inclusion criteria | DMARD discontinuation | DMARD continuation | Risk ratio for disease relapse (95% CI) | ||
|---|---|---|---|---|---|---|---|
| N relapses at 1 year/N patients | Relapse rate at 1 year | N relapses at 1 year/N patients | Relapse rate at 1 year | ||||
| Nonrandomized clinical studies | |||||||
| Brocq 2009 | IFX, ADA, ETN | DAS28 ≤ 2.6 For >6 months | 15/20 | 75% | – | – | – |
| Tanaka 2010 | IFX | DAS28 ≤ 3.2 For >6 months | 45/102 | 45% | – | – | – |
| Tanaka 2013 | ADA | DAS28 ≤ 2.6 For >6 months | 27/52 a | 52% | 4/23 a | 17% | |
| Takeuchi 2015 | ABA | DAS28 CRP ≤ 2.3 For >6 months | 22/34 | 65% | – | – | – |
| Aguilar-Lozano 2013 | TCZ | DAS28 ≤ 2.6 For >6 months | 25/45 | 56% | – | – | – |
| Nishimoto 2014 | TCZ | DAS28 ≤ 3.2 For >6 months | 162/187 | 87% | – | – | – |
| Huizinga 2014 | TCZ c | DAS28 < 2.6 For >3 months | Add-on: 229/277 Switch: 245/277 | 83% 89% | – | – | – |
| Randomized controlled trials | |||||||
| Gotzsche, 1996 | MTX | Stable DMARD | 31/55 | 56% | 8/57 | 14% | 4.0 (2.0, 7.7) |
| ten Wolde 1996 | MTX | Good long-term response | 53/143 | 35% | 30/142 | 21% | 1.8 (1.2, 2.6) |
| Chatzidionysiou 2013 | ADA | DAS28 ≤ 2.6 For >3 months | 12/15 b | 80% | 8/16 b | 50% | 2.8 (1.4, 5.9) |
| Smolen 2013 | ETN | DAS28 ≤ 3.2 For >6 months | 113/197 | 57% | 35/201 | 17% | 2.3 (1.8, 2.8) |
| van Vollenhoven 2015 | ETN | DAS28 ≤ 3.2 For >6 months | 20/23 | 87% | 11/23 | 48% | 4.0 (1.3, 12.5) |
a Nonrandomized control group (patients who refused ADA discontinuation).
c TCZ as add-on to MTX or switch to MTX in MTX inadequate responders.
For bDMARDs, the potential interest of the association of MTX to maintain bDMARD-free remission was explored in one study, corresponding to the second-year extension of the ACT-RAY trial, which tested tocilizumab (TCZ) in addition (add-on) or replacing MTX in established RA patients inadequately responding to MTX . The relapse rate when tapering was numerically, although not statistically significantly, higher – 83% versus 88% for the patients in the add-on group versus those in the switch group – with a median time to flare following TCZ discontinuation of 113 days in the add-on group versus 84 in the switch group. Importantly, sustained drug-free remission with either strategy was apparently quite limited.
Several studies have assessed the impact of tapering/withdrawal on radiographic structural damage progression. In observational studies, in which bDMARDs were restarted as soon as flare/relapse was observed, no significant progression was observed in the discontinuation group , also with no significant difference compared with the continuation group . However, in the PRESERVE trial , the discontinuation of ETN was maintained up to the 1-year follow-up visit, and a significant difference was observed between the continuation and discontinuation groups, although the 1-year change in van de Heijde-modified Sharp (vdHS) score was quite small.
Importantly, these studies found no difference in terms of safety. More specifically, the risk of infection and serious infection was not reduced, although this risk has been thought to be dose-dependent . Again, this observation might be explained by a phenomenon called “dilution of the susceptible” or “healthy survivor bias”: patients who are the most at risk of infections are likely to stop their DMARDs rapidly after treatment initiation, and they are less likely to be part of the population in stable and sustained remission with such therapies .
Importantly, despite the risk of relapse, even when the treatment is maintained, some patients did not relapse during the 1-year follow-up. Although no plateau was reached at 1 year in these studies, indicating that the probability of relapse continued to increase with time after DMARD discontinuation, prolonged DMARD-free remission seemed possible for a subset of RA patients ( Table 1 ) in “sustained remission.” Exploratory analyses attempted to identify predictors of sustained DMARD-free remission. In the remission induction by Remicade in RA (RRR) and HONOR studies , a significant correlation was found between DAS28 at the time of bDMARD discontinuation and the 1-year risk of relapse, with a threshold value of 2.1 or 2.2. Interestingly, such a DAS28 value was also reported as the value under which rheumatologists considered RA as appropriately controlled in a previous qualitative study of therapeutic target . However, there was a large overlap in terms of initial DAS28 between the two groups – relapse versus no relapse ( Fig. 2 ) – which indicates that the 2.1–2.2 threshold, although potentially relevant from a clinical point of view, was not applicable at the individual level to identify patients with no or low risk of relapse in the months following DMARD discontinuation. In another study of TCZ, the authors identified serum interleukin 6 (IL-6) and metalloproteinase 3 (MMP3) levels as predictors of relapse after TCZ discontinuation . These results still need to be confirmed.
DMARD tapering in RA
Half-dose reduction
If DMARD discontinuation appears at a high risk of relapse, DMARD tapering could be an interesting alternative. Such tapering strategies have been suggested by some authors who argue that the dose of biologics given in clinical practice may be unnecessarily high . A first observational study assessed progressive down-titration of IFX over 1 year . Down-titration was feasible for 45% of patients, with a mean dose reduction of 60% (mean dose of 224 mg at baseline to 130 mg after 1 year) ( Table 2 ). Discontinuation of IFX was possible for 16% of patients. In this study, which included RA patients with long disease duration, who received IFX in the early bDMARD era, the risk of relapse was high – 48–100% – when the dose was reduced by 50% or more.
| Reference | Drug | Inclusion criteria | Tapering strategy | DMARD dose tapering | DMARD continuation at full dose | Risk ratio for disease relapse (95% CI) | ||
|---|---|---|---|---|---|---|---|---|
| N relapses /N pts | Relapse rate | N relapses /N pts | Relapse rate | |||||
| Nonrandomized clinical studies | ||||||||
| van der Maas 2012 | IFX | DAS28 ≤ 3.2 For >6 months | Dose reduction IFX by 25, 50, 75 or 100% Time line: 12 mo. | 25%: na/51 50%: na/39 75%: na/57 100%: na/25 | 25%: ~25% a 50%: ~48% a 75%: ~69% a 100%: ~75% a | – | – | – |
| Randomized controlled trials | ||||||||
| Smolen 2013 | ETN | DAS28 ≤ 3.2 For >6 months | Half-dose ETN Time line: 12 mo. | 42/201 | 21% | 35/201 | 17% | |
| van Vollenhoven 2015 | ETN | DAS28 ≤ 3.2 For >6 months | Half-dose ETN Time line: 12 mo. | 15/27 | 56% | 11/23 | 48% | HR 1.0 (0.4, 2.4) |
| van Herwaarden 2015 | ETN ADA | DAS28 ≤ 2.6 For >6 months | Injection spacing with bDMARD discontinuation at third step Time line: 18 mo. | 88/121 | 73% | 16/59 | 27% | |
| Fautrel 2015 | ETN ADA | DAS28 ≤ 2.6 For >6 months | Injection spacing with bDMARD discontinuation at fourth step Time line: 18 mo. | 49/64 | 77% | 34/73 | 47% | HR 2.4 (1.5, 3.8) |
| Haschka 2015 | Any csDMARD Any bDMARD | DAS28 < 2.6 For >6 months | Dose reduction by 50% or 50 then 100% Time line: 12 mo. | 50%: 14/36 50/100%: 14/27 | 50%: 39% 50/100%: 52% | 6/38 | 16% | |
a Extrapolated from Fig. 1 of the paper, follow-up ranging from 9 to 12 months.
More recent studies are probably more relevant to address the question of DMARD tapering. Actually, five RCTs are now available, comparing tapering strategies to full-dose continuation ( Table 2 ). Two were already mentioned trials that used a superiority design to demonstrate the benefit of ETN continuation – either full-dose 50 mg/week or half-dose 25 mg/week – compared with its discontinuation, in established RA patients who achieved sustained LDA with ETN and MTX. Thus, the risk of relapse in the two dosage arms could be assessed. In PRESERVE, the risk of relapse, defined as DAS28 > 3.2 with a DAS28 change of >0.6, was similar with ETN at full and half doses, about 40% at 1-year follow-up ( Fig. 3 ) . In DOSERA, the risk of relapse, based on the same definition, was slightly higher, about 50%, with no statistical differences between the two dosages .
Progressive and disease-activity-driven tapering
Progressive spacing of bDMARD injections is an alternate way to reduce bDMARD exposure, with the substantial advantage of being (1) implementable for all bDMARDs regardless of the availability of specific half-dose packaging, and (2) favored by patients who are often bothered by the need for recurrent injections. Such a strategy was tested in two equivalence trials comparing a 3-monthly progressive spacing of ETN or ADA injections to their continuation at full dose in established RA patients in DAS28 remission . In the Spacing of TNF-blocker injections in Rheumatoid Arthritis Study (STRASS) trial , the time between two subcutaneous bDMARD injections was progressively increased at each 3-month visit if remission was maintained, up to complete discontinuation at the fourth step ( Fig. 4 ). In case of relapse, bDMARD injections were reset to the previous injection interval. The primary criterion was the disease activity over the 18-month follow-up, estimated by repeated DAS28 measures. The conclusions of the study are weakened by the trial enrolling fewer patients than expected. However, the main message was that equivalence was not demonstrated in the STRASS trial because the standardized difference in terms of disease activity was 19% (95% CI: −5%, 46%), the upper limit of the confidence interval being outside of the prespecified equivalence interval (−30%, 30%). In addition, RA relapses occurred more frequently: 77% and 47% in the tapering versus continuation arms ( p = 0.004), with a median time to relapse of 9 and 18 months, respectively. Importantly, the mean DAS28 at relapse was in the moderate activity range – 3.5 and 3.1, respectively – so the relapse did not correspond to full loss of response to the bDMARD. In case of relapse, the reversal of injection spacing allowed for achieving a new response in 88% of patients. Thus, the structural damage progression was similar between the two groups. In addition, findings did not differ between ADA- and ETN-treated patients or between patients treated with bDMARDs as monotherapy or combined with csDMARDs. An important finding of the STRASS trial was the possibility to implement the step-down strategy in most of the patients – 48/64 (75%) – with, at the end of the study, 9% at step 1, 11% at step 2, 16% at step 3, and 39% at step 4 (bDMARD full stop). Only a few patients from the spacing arm received an ETN or an ADA full-dose regimen at the end of the study. Of course, additional long-term information is needed to fully appreciate the sustainability of such a step-down strategy. However, no clear predictor of relapse or failure to dose tapering was identified.
The DRESS trial used a somewhat comparable tapering design with progressive spacing of ADA or ETN subcutaneous injections up to discontinuation at the third step. It included established RA patients in DAS28 LDA . However, the primary criteria to conclude the equivalence of the spacing and the full-dose continuation strategies were “persistent flare,” defined as disease flare (i.e., change in DAS28 of >1.2 or change in DAS28 of >0.6 with DAS28 of >3.2) not recovered at the next visit 3 months later, which differed substantially from the previous study, but was fully relevant from a clinical point of view. Although the risk of flare was higher in the spacing than in the continuation arms – 73% versus 27% ( p < 0.001) – the risk of persistent flare was equivalent: 14/119 (12%) versus 5/50 (10%), corresponding to a difference of 2% (95% CI: −12%, 12%), within the predefined equivalence interval (−20%, 20%). No difference was observed between ADA and ETN treatment ( Fig. 5 ). In addition, the mean disease activity at study end did not differ between the two groups, but fewer patients were in the American College of Rheumatology (ACR)/EULAR Boolean remission at the end of follow-up in the spacing versus the continuation arm (24% vs. 41%, p = 0.02). Moreover, structural damage progression was significantly higher in the spacing arm, with a mean difference in vdHS score of 0.6 (95% CI: 0.17, 1.0), with no difference in erosion score, but a difference in the joint-narrowing vdHS score, and a higher proportion of progressors (ΔvdHS > 0.5) (i.e., 32% vs. 15%). These differences are small and irrelevant from a clinical point of view; however, when these differences would increase in following years, they would become clinically relevant. Notably, in case of flare, patients were proposed to receive intramuscular or intra-articular steroid injections, which might have led to an adequate prevention of persistent flare without the same benefit on structural damage. Additional information is probably needed with regard to this specific point. Finally, as in the previous study, bDMARD reescalation was successful in most patients when a flare occurred, and no difference was observed in terms of safety events.
Finally, another trial, RETRO, tested a more global approach to DMARD tapering . In this 1-year study, both csDMARD- and bDMARD-treated patients with established RA were included after having achieved DAS28 remission for >6 months. Patients were randomized in three groups: (1) full-dose continuation of all DMARDs; (2) 50% tapering of all DMARDs, either by dose reduction or by injection spacing; and (3) 50% than 100% (if remission was maintained at 6 months) tapering of all DMARDs. As expected, the more aggressive the tapering strategy, the greater the risk of relapse: 16% in arm 1, 39% in arm 2 ( p = 0.03), and 52% in arm 3 ( p = 0.003). Interestingly, although patients with positive autoantibodies were at a greater risk of relapse, no difference was observed between bDMARD and csDMARD patients or between patients in ACR/EULAR Boolean remission at study entry. At the end of follow-up, the number of patients in remission or LDA was numerically, although not significantly, lower in the tapering than in the continuation: 6%, 16%, and 15% in arms 1, 2, and 3, respectively. No structural damage assessment was available in this study. Although the population was heterogeneous in terms of DMARDs and the change of all DMARDs at the same time made interpretation more complex, the RETRO trial confirmed the feasibility of DMARD tapering for a relevant proportion of patients.
The “how”: possible strategies for tapering of RA therapies
Recent evidence for DMARD discontinuation in RA
DMARD discontinuation has been regularly tested in established RA patients in both observational studies and randomized controlled trials (RCTs) ( Table 1 ). However, substantial methodological heterogeneity was noted: enrolled patients had stable disease, defined only qualitatively in the oldest publications, or based on DAS28 LDA/remission in the most recent ones. In the latter publications, patients with DAS28 remission or LDA were selected, with a mean remission or an LDA duration of usually at least 6 months. The definition of flares or relapses differed across studies: recurrence of one or more joints with clinical synovitis, loss of DAS remission or LDA, loss of remission or LDA associated with DAS28 change of >0.6, or 1.2. Despite this heterogeneity, several observations are possible. First, observational studies were mainly conducted in patients treated with biologics in routine or in long-term extension (LTE) following industry-sponsored RCTs . The relapse rate at 1 year after bDMARD discontinuation ranged from 45% to 88%, whatever the bDMARD considered ( Table 1 ). Second, whatever the DMARD, methotrexate (MTX) or bDMARD, the risk of disease-activity exacerbation in the year following DMARD discontinuation was high, >50% except in two studies . In RCTs, comparison with the DMARD continuation arm clearly identified an excess risk of relapse, with risk ratios of ≥2 ( Table 1 ). Of note, this risk did not differ between MTX and bDMARDs, or between bDMARDs themselves. The reinitiation of the DMARD – mainly studied in bDMARD studies – was associated with an adequate new response in most patients who relapsed . Interestingly, the relapse rate when the DMARD was continued at the same dose was sometimes high, reaching 50% in two studies . One might interpret these results in several ways: (1) a residual and unavoidable risk of RA flare whatever the remission duration as part of RA natural history, which is incorrectly attributed to the tapering attempt; (2) the persistence of spontaneously resolving disease-activity fluctuations – thus not requiring DMARD adaptation or change – as was shown in previous cohort studies ; (3) the lack of reliable flare or relapse definition enabling the discrimination between the first two points ; and (4) a possible “nocebo” effect, making flare more expected when the patient is told that treatment is going to be reduced (though not in placebo-controlled studies, however).
Related posts:
Treating to target in established rheumatoid arthritis: Challenges and opportunities in an era of novel targeted therapies and biosimilars
How does established rheumatoid arthritis develop, and are there possibilities for prevention?
Monitoring in established RA: Role of imaging and soluble biomarkers
Evaluating and mitigating fracture risk in established rheumatoid arthritis
Patient-centred care in established rheumatoid arthritis
Quality in rheumatoid arthritis care
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
