Article Text

Download PDFPDF

Ability of disease-modifying antirheumatic drugs to prevent or delay rheumatoid arthritis onset: a systematic literature review and meta-analysis
  1. Stéphane Hilliquin1,2,
  2. Benjamin Hugues1,2,
  3. Stéphane Mitrovic1,2,
  4. Laure Gossec1,2,
  5. Bruno Fautrel1,2
  1. 1 Rheumatology, UPMC, Institut Pierre Louis d'épidémiologie et Santé publique, GRC 08, Paris, France
  2. 2 Department of Rhumatologie, AP-HP, GH Pitié Salpêtrière, Paris, France
  1. Correspondence to Prof Bruno Fautrel, UPMC, Institut Pierre Louis d'épidémiologie et Santé publique, GRC 08, Paris 75646, France; bruno.fautrel{at}


Background Recent advances in knowledge of the pathogenesis of rheumatoid arthritis (RA) has led to promoting very early intervention.

Objectives To assess the efficacy of therapeutic interventions in preventing or delaying RA onset with a systematic literature review (SLR) and meta-analysis (MA).

Methods The SLR aimed to include all reports of randomised controlled trials of disease-modifying antirheumatic drugs or glucocorticoids used in patients presenting genetic and/or environmental risk factors for RA and/or systemic autoimmunity associated with RA, and/or symptoms without clinical arthritis and/or unclassified arthritis and in patients with RA. We searched PubMed, EMBASE and Cochrane databases for English articles published from 2006 to 2016 using the keywords ‘undifferentiated arthritis’ or ‘very early rheumatoid arthritis’ with ‘therapy’ or ‘treatment’. Main outcome was RA occurrence, defined as fulfilment of the 1987 ACR criteria. The MA was performed with RevMan with the Mantel-Haenszel method.

Results Among 595 abstracts screened, 10 reports of trials were selected. The studies included 1156 patients, with mean symptom duration 16.2±12.6 weeks. The occurrence of RA was available for nine studies, assessing methylprednisolone, methotrexate, a tumour necrosis factor blocker, abatacept or rituximab. In the group arthralgia without arthritis (people at risk of RA), the MA of the two available studies did not show significant reduction in RA occurrence at week 52 or more (pooled OR 0.74, 95% CI 0.37 to 1.49). For people with undifferentiated arthritis, the MA of the seven available studies revealed significant risk reduction with OR 0.73(95% CI 0.56 to 0.97).

Conclusions This MA demonstrates that early therapeutic intervention may significantly reduce the risk of RA onset in this very first phase of the disease.

  • rheumatoid arthritis
  • treatment
  • corticosteroids
  • dmards (synthetic)
  • dmards (biologic)

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.


During the last decades, substantial knowledge has accumulated on the very early stages of rheumatoid arthritis (RA), notably the very early immunological pathogenic mechanisms leading to RA.1–4 This knowledge has deeply altered the nosological RA concept and its diagnosis.

For a long time, RA diagnosis required a quite complete and comprehensive clinical presentation, including bilateral symmetrical polyarthritis, involving the hands, eventually associated with serum rheumatoid factor (RF), nodules or radiographic joint erosions, as included in the 1987 American College of Rheumatology (ACR) classification criteria.5 This ‘full presentation’ does not fit with RA early stages, and the 1987 ACR criteria were found to adequately classify patients as having RA only 2 years after disease onset.6 In 2010, the joint effort of the ACR and European League Against Rheumatism (EULAR) widened the spectrum of early RA by reducing the minimal synovitis number to 1 and including serum anticitrullinated peptide antibody (ACPA) positivity in addition to RF as immunological biomarkers.7 The new classification showed higher sensitivity to detect patients with early RA8 and affected the diagnostic concept of unclassified arthritis (UA) as well as very early RA (VeRA).9 10

The combination of advances in RA pathogenesis and progress in RA diagnosis has contributed to redefining the RA early stages as a continuum spreading over several years4 11 starting from (1) a first autoimmune phenomenon related to a host–environment interaction (eg, interaction between smoking and the presence of the shared epitope leading to ACPA production); (2) preclinical RA (pre-RA), in which levels of autoimmunity biomarkers increase and mature, potentially associated with mild inflammatory features (eg, arthralgia without arthritis/synovitis); (3) UA with at least one synovitis present, without satisfaction of the 1987 ACR criteria (but potentially satisfying the 2010 ACR/EULAR criteria); and finally (4) defined RA, with ‘full-picture’ RA and satisfaction of 1987 ACR classification criteria.10 11 These concepts have been retained in recent EULAR recommendations for research of individuals at risk of RA.12

Besides diagnosis, the therapeutic issue is also important. Early therapeutic interventions, within the first months after RA onset, were clearly found to be associated with better RA outcomes,13–15 thereby validating the concept of a ‘window of opportunity’. In addition, the PROMPT trial demonstrated the ability of early methotrexate initiation to prevent onset of RA in patients with UA.16 17 This situation raised the question of delaying or preventing RA if RA treatments are started at preclinical or in the very early clinical stages of the disease.11

Although international clinical practice guildelines focus on methotrexate or other conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) in early RA,18 numerous other therapeutic options are available, and several, including glucocorticoids or biologic DMARDs (bDMARDs), have been tested in individuals at risk of RA. A recent meta-analysis (MA) of studies of experimental animal models suggested that DMARDs are not equally efficacious in the prevention or treatment of the early arthritis animal model.19

Thus, we conducted a systematic literature review (SLR) and MA of randomised controlled trials (RCTs) of patients at risk of RA to assess the efficacy of glucocorticoids, csDMARDs or bDMARDs for preventing or delaying RA development and/or blocking structural damage. The notion of prevention of RA refers to the ability of a treatment to block the pathogenic process and prevent more established forms of RA. Thus, the target population for such an action is people at risk of RA (family history and presence of (high titre) autoantibodies, or with arthralgia and autoantibodies, or people with UA).


Search strategy

We performed a systematic literature search of PubMed, Medline, EMBASE and the Cochrane Controlled Trials Register up to June 2017 for articles published from 2006 to 2016 and EULAR and ACR scientific meeting abstracts from the last 2 years (2015 and 2016). We used the following key words: ‘Arthritis, Rheumatoid’(MeSH) AND ‘very early’ AND ‘treatments’(all fields) OR ‘therapy’(all field), or ‘undifferentiated arthritis’(All Fields) AND ‘therapy’(all fields) or ‘treatments’(all fields). We limited our search to English-language reports of RCTs of adults ≥18 years old. In addition, we hand-searched reference lists of papers initially detected to identify additional relevant reports. The reports of clinical trials were initially selected on the basis of the title and abstract, then the full text. Duplicate references were removed.

Study selection criteria

To be selected, reports had to satisfy the following:

  1. The study design should be an RCT.

  2. The enrolled patient diagnosis should be one of (A) patients presenting genetic and/or environmental risk factors for RA and/or systemic autoimmunity associated with RA, and/or symptoms without clinical arthritis and/or UA7 12 20; (B) patients with clinical arthritis evolving for <16 weeks and fulfilling the 2010 ACR/EULAR criteria but not the 1987 ACR criteria.

  3. Study treatment should be glucocorticoids or any other DMARD, either a csDMARD (methotrexate) or bDMARD (tumour necrosis factor (TNF) blocker or other mode of action).

  4. Study outcomes should be measured at week 52 or closest time point.

Data extraction

Two independent readers (SH, BH) extracted the following data by using a standardised form: patient characteristics at baseline (ie, demographics and disease characteristics, including classification criteria fulfilment); therapeutic intervention; occurrence of RA at week 52 or closest time point, defined as fulfilment of the 1987 ACR classification criteria17 21–25 or ACR EULAR 2010 or according to the rheumatologist’s opinion21 26; clinical remission rates at week 52, defined by validated composite criteria (ie, Disease Activity Score in 28 joints (DAS28),17 21–25 27 Simple Disease Activity Index (SDAI), Clinical DAI (CDAI) or boolean,25 with adequate threshold); structural damage progression seen on X-rays at week 52 (table 1) based on the van der Heijde-modified Sharp score or any other validated score17 21 23–25 27 and safety based on a descriptive analysis.

Table 1

Main outcome results

To define specific phases of RA, patients were classified into groups according to the EULAR recommendations for terminology12 as follows: (a) genetic risk factors for RA; (b) environmental risk factors for RA; (c): systemic autoimmunity associated with RA; (d) symptoms without clinical arthritis; (e) UAs; (f) RA.

Study quality was assessed by the Jadad scale28 with two questions (answer Yes/No) for randomisation, two for masking, and one (answer Yes/No) evaluating the reporting of withdrawals and dropouts. A total of 5 points could be awarded, with higher scores indicating higher quality.

Statistical analysis and MA

The MA was performed accordingly to the Cochrane Collaboration guidelines29 for RA occurrence, defined as a definite RA, which is mostly according to the 1987 ACR classification criteria in the literature (or ACR EULAR 2010) at week 52 or closest time point, radiographic progression and clinical remission. Concerning RA occurrence, data at week 52 and beyond this time were pooled to strengthen the results. A sensitivity analysis was conducted to isolate the impact of TNF-blocker treatment. Statistical heterogeneity was tested by the χ2 Q test30; with significant heterogeneity, a random-effects model was used. The MA was performed with RevMan V.5.3 (The Nordic Cochrane Centre, Copenhagen, 2014) with the Mantel-Haenszel method, estimating ORs and 95% CIs. A descriptive analysis was performed for other measures such as the DAS28, Health Assessment Questionnaire (HAQ) and side effects (infectious and intolerance).


Selected studies

The search identified 595 abstracts, with reports of 10 RCTs selected (including 2 congress abstracts) and 10 exploited for analysis (figure 1). The main reasons for exclusion were disease duration at baseline (many studies included patients with RA evolving for >16 weeks), study design (ie, non-RCTs), study endpoints different from the outcomes of interest and incomplete results (ie, missing data for means and/or SD). The mean Jadad score was 5, which indicates high methodological quality of studies.

Figure 1

Flow of studies in the review. ACR, American College of Rheumatology; RCT, randomised controlled trial.

Seven were related to (e) criteria, two to (d) and one to (f).12 The therapeutic strategies tested were methylprednisolone at a single dose of 8021 or 120 mg,22 intramuscularly in two studies; dexamethasone at the dose of 100 mg IM at week 0 and week 631 oral methotrexate up to 30 mg/week for X weeks or months in one study17; TNF blockers—infliximab (3 mg/kg at weeks 0-2-4-6−14+/-22)23 32 or etanercept (50 mg/kg/wk)25 27 at labelled doses—in four studies, used alone or with methotrexate (up to 30 mg/wk); intravenous abatacept (100 mg/kg every 2 weeks for 1 month, then monthly)24 at a labelled dose in one study; and finally intravenous rituximab at 1 g once only in one study (table 2).

Table 2

Study characteristics

The SLR and MA included 1239 patients (mean percentage of women 66.0% with weighted mean age 45.8±15.2 years and mean symptom duration 16.2±12.6 weeks.

Data synthesis

Preventing or delaying RA occurrence

RA occurrence, defined as satisfaction of the 1987 ACR classification criteria, was found in 9 of 10 papers assessing methylprednisolone (80 to 120 mg intramuscularly), dexamethasone, methotrexate, TNF blocker (infliximab in the Saleem and Durez trial; etanercept in EMPIRE), abatacept or rituximab.

Two studies are related to arthralgia without arthritis (d)26 31 evaluating dexamethasone and rituximab. seven are related to UA (e).21–25 32 33

In the group arthralgia without arthritis (d), the MA of the two available studies did not show significant reduction in RA occurrence at week 52 or more (pooled OR 0.74, 95% CI 0.37 to 1,49) (figure 2A).

Figure 2

(A) RA diagnosis at week 52 or more including tumour necrosis factor (TNF) blockers. (B) RA diagnosis at week 52 or more not including TNF blockers. ACPA, anti-citrullinated protein antibody; MethylPDN, methylprednisolone;  MTX, methotrexate; PDN, prednisolone; RA, rheumatoid arthritis; RF, rheumatoid factor; TNF, tumour necrosis factor; UA, undifferentiated  arthritis; VeRA, very early rheumatoid arthritis.

For people with undifferentiated arthritis (e), the MA of the seven available studies revealed significant risk reduction with OR 0.73 (95% CI 0.56 to 0.97). All drugs tended to reduce the risk of RA occurrence, except TNF blockers (figure 2A).

As a sensitivity analysis, the MA was performed without the 2 TNF-blocker studies, which resulted in a more significant pooled OR 0.68 (95% CI 0.50 to 0.92) (figure 2B).

Clinical remission

Clinical remission at week 52, according to the DAS28, SDAI, CDAI or boolean definitions, was available for five studies of glucocorticoids and TNF blockers (etanercept or infliximab) or abatacept. The Saleem and Durez study used another criteria (no swollen joint and C reactive protein level <10 mg/L). Only the COMET trial27 demonstrated a significant effect of etanercept on remission. The MA revealed that early intervention increased the odds of achieving remission (pooled OR 1.84, 95% CI 1.08 to 3.16) (figure 3).

Figure 3

Clinical remission at week 52. ACPA, anti-citrullinated protein antibody; MethylPDN, methylprednisolone; MTX, methotrexate; PDN, prednisolone; RA, rheumatoid arthritis; RF, rheumatoid factor; UA, undifferentiated  arthritis; VeRA, very early rheumatoid arthritis.

Radiographic progression

Data on radiographic progression were available for five studies, evaluating methylprednisone, methotrexate or a TNF blocker (etanercept and infliximab). The outcomes were the Sharp score (modified or not) and Larsen score. No significant risk reduction was revealed for radiographic progression (figure 4). The MA yielded a pooled OR of 1.36 (95% CI 0.82 to 2.27). We found no difference between treatments for radiographic progression. The analysis without TNF blockers did not alter the results.

Figure 4

Absence of radiographic progression at week 52. ACPA, anti-citrullinated protein antibody; MethylPDN, methylprednisolone; MTX, methotrexate; RA, rheumatoid arthritis; RF, rheumatoid factor; UA, undifferentiated  arthritis; VeRA, very early rheumatoid arthritis.

Other outcomes and side effects

With the descriptive analysis, similar side effects were observed between placebo and methylprednisone without notable difference.21 22 The observed side effects were the expected ones: hypertension, lower limb oedema,22 anaphylactic reaction and mood swings.21 The PROMPT study found no significant safety difference for methotrexate versus placebo (26/55 with methotrexate and 18/55 with placebo, p=0.17). Side effects described were benign gastrointestinal events, elevated serum liver enzyme levels and dermal/mucosal events with methotrexate. bDMARDs were associated with respiratory and urinary tract infections, with two severe cases.25 No malignancy was identified. There were 10 safety events with abatacept versus 11 with placebo.24 The most frequently reported events were nasopharyngitis, urinary tract infection and gastroenteritis. The abstract for the PRAIRI study (rituximab) did not specify side effects.26


The present SLR and MA provide information that a very early therapeutic intervention may significantly reduce the risk of RA onset with patients at risk of RA and significantly increase the rate of clinical remission. Although the notion of a window of opportunity is well accepted in people with already diagnosed RA,15 34 our work reveals that an even earlier therapeutic intervention could prevent RA or delayed its onset. This conclusion seems to be available in patients with arthritis. In symptomatic patients without arthritis, no significant reduction was observed, potentially partly due to a lack of power with the only two available studies.

This work also suggests that the beneficial effect of very early treatment in RA could differ between csDMARDs or bDMARDs. Although a reduced risk of RA occurrence was observed with glucocorticoids, methotrexate, abatacept or rituximab, the trend seemed not confirmed for TNF blockers.23 25 This finding is likely to be a class effect rather than a single molecule effect because it was observed with two different agents, one soluble receptor (etanercept) and one monoclonal antibody (infliximab). They may be not as important in the very early steps of the disease in which autoimmune phenomenon are present but not joint or systemic inflammation.35 TNF blockers have been mainly tested, and are currently recommended, in established active RA and eventually in early RA with pejorative prognostic factors such as high swollen joint count, increased acute-phase reactant levels or joint erosions; this is in line with the association between TNF and detectable inflammation.36–40 Pharmacological agents such as glucocorticoids, methotrexate, rituximab or abatacept may have a broader effect and act higher in the pathogenic cascade, including antigen presentation and early steps of the autoimmune reaction; they could thus prevent the immune system activation, whereas TNF blockers could only reduce already existing inflammation. We have no data available for interleukin 6 (IL-6) blocking agents, although IL-6 seems to be involved in the very early steps of RA pathogenesis.41

Several strengths of the study must be underlined. The study applied the methodological standards recommended by the Cochrane collaboration for an SLR and MA, including double data extraction and entry.29 Although people in very early phases of the disease constitute a challenging population for clinical research, nine reports of RCTs were identified and the data could be integrated in the MA. These trials cover almost all possible modes of action for RA, except IL-6 blockers and JAK inhibitors.

However, our work has some limitations. Although data about about RA diagnosis according to 1987 ACR classification criteria were available in most of the studies (seven of nine), only a few reported data on clinical remission (five of nine). Structural damage information was assessable in five studies; however, the progression was small in RA, and we could not identify any significant benefit of early therapeutic intervention for this outcome. In addition, we found substantial heterogeneity in the outcome measures used in the selected trials: DAS28, SDAI, CDAI or Boolean definitions for remission and Larsen or van der Heijde-modified Sharp score for radiographic progression. For feasibility reasons, we pooled the remission rates or percentages for patients without structural damage progression, regardless of the tool used. This move could have biased our results in part.

An important concern comes with the distinction of the very early steps of RA.12 35 We chose the cut-off of 4 months of disease duration to select the studies for our MA, which was based on data from a few studies25 27 using the cut-off of 3 to 4 months to define the very early RA phase. This choice is of course partly arbitrary and reveals the complexity to define the initial RA phases, which constitute a continuum1 rather than a succession of clearly different health states.35 42 The 2010 ACR/EULAR criteria7 allowed for identifying patients with RA at an earlier stage than did the 1987 criteria.8 However, the criteria are operational only for patients with significant RA symptoms and do not cover the whole spectrum of people with less specific symptoms such as those with arthralgia or limited arthritis, with or without family history of RA, and/or with or without serum ACPA positivity. The development of clinical practice guidelines for people at risk of RA was an important step forward12 but did not completely fix the overlap of the existing definitions of the RA early phases.35 42

Despite these difficulties, our results reinforce the view of ‘the sooner the better’ in terms of therapeutic decision-making within the pathogenic RA continuum. This paradigm raises an important additional question related to the duration of such a very early therapeutic intervention aiming to prevent RA onset or completely abate the disease. Whatever the RA stage, the risk of relapse seems substantial when treatments are not maintained.17 21–26 However, there is potential for ‘immunological remission’ in some patients with RA (ie, resolution of any sign of joint or systemic inflammation with disappearance of serum autoantibodies (RF or ACPA)).43 Early or very early intervention may favour such immunological remission and could correspond to some kind of resetting of the immune system with complete resolution of any autoimmune phenomenon. The optimal strategy for such patients could then be an induction therapeutic sequence to prevent RA or achieve immunological remission, then a drug tapering or discontinuation sequence to reach sustained and stable drug-free and disease-free states.44–47 This move would probably require as an intermediary step a better assessment and quantification of the risk of developing RA for a given patient to facilitate the implementation of more personalised therapeutic schemes. A risk stratification score, based on family history as well as patient clinical and biological features, has been proposed in the context of the Leiden Early Arthritis Clinic.47 Finally, this notion of prevention in patients at risk of developing RA needs to be handled with caution for two reasons. First, in all the trial, the treatment of patients with arthralgia and autoantibodies or with UA have mainly shown their capacity to delay or postpone RA onset, but only a minority of them will durably remain asymptomatic if the tested DMARD is discontinued.16 17 26 Second, it must be kept in mind that these patients may correspond to patients achieving spontaneous remission with no or only little role for the DMARD. In a recent work conducted in the ESPOIR and the Leiden early arthritis cohorts, such an evolution—that is, DMARD-free sustained remission—could be observed in 5.4% (29/533 in ESPOIR) to 11.5% (85/738 in LEAC).15 It is important to note that delay is not prevention. It remains to date unknown whether a minority will remain asymptomatic after DMARD is discontinued.

This question deserves to be further studied, why not by creating a two arms study with early intervention in case of arthralgia and autoimmunity comparing intervention only when clinical arthritis develops.

In conclusion, these SLR and MA clearly demonstrate the potential benefits of very early therapeutic intervention for people who start RA and specifically its ability to prevent established RA. Our results fit perfectly with the 2017 EULAR campaign on early actions in rheumatic disorders: ‘Don’t delay, Connect today’ (



  • SH and BH contributed equally.

  • Handling editor Josef S Smolen

  • Contributors BF had the initiative for drafting this article. The literature search and writing of the article were performed equally by BH and SH. The data layout, meta-analysis and rereading was carried out by SM. The overall proofreading was achieved by both BF and LG.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval Ethics Committee/Institutional Review Board approval obtained.

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