Article Text

This article has a correction. Please see:


Extended report
The Stop Arthritis Very Early (SAVE) trial, an international multicentre, randomised, double-blind, placebo-controlled trial on glucocorticoids in very early arthritis
  1. Klaus P Machold1,
  2. Robert Landewé2,
  3. Josef S Smolen1,
  4. Tanja A Stamm1,
  5. Désirée M van der Heijde3,
  6. Kirsten N Verpoort3,
  7. Kerstin Brickmann4,
  8. Janitzia Vázquez-Mellado5,
  9. Dimitri E Karateev6,
  10. Ferdinand C Breedveld3,
  11. Paul Emery7,
  12. Thomas W J Huizinga3
  1. 1Medical University of Vienna, Vienna, Austria
  2. 2University of Maastricht, Maastricht, The Netherlands
  3. 3Leiden University Medical Centre, Leiden, The Netherlands
  4. 4Medical University of Graz, Graz, Austria
  5. 5National Autonomous University of Mexico, Mexico City, Mexico
  6. 6State Institute of Rheumatology, Moscow, Russian Republic
  7. 7Academic Unit of Musculoskeletal Disease, Leeds, UK
  1. Correspondence to Klaus P Machold, Medical University of Vienna, Währinger Gürtel 18-20, Vienna A-1090, Austria; klaus.machold{at}


Background Glucocorticoids (GCs) are often used as early arthritis treatment and it has been suggested that they induce remission or at least delay the development of rheumatoid arthritis (RA) and the need to start disease-modifying antirheumatic drugs (DMARDs).

Objective To test the effect of GCs on patients with very early arthritis (symptom duration of <16 weeks) in a randomised controlled trial.

Methods Patients received a single intramuscular injection of 120 mg methylprednisolone or placebo (PL) and were followed up for 52 weeks. Primary end point was drug-free clinical remission, both at weeks 12 and 52. Among secondary outcomes were fulfilment of remission criteria at weeks 2, 12 or 52, time course of ‘core set variables’ and proportion of patients starting DMARDs.

Results 17.0% of all analysed subjects (65/383) achieved persistent remission: 17.8% (33/185) of the PL group, 16.2% (32/198) of the patients receiving methylprednisolone (OR=1.13, 95% CI 0.66 to 1.92, p=0.6847). Analyses of secondary end points showed significant clinical benefits of the GC only at week 2. These differences subsequently disappeared. DMARDs were started in 162 patients: 50.3% methylprednisolone and 56.7% PL patients had to start DMARD treatment (OR=0.78, 95% CI 0.49 to 1.22, p=0.30). Significantly more patients with polyarthritis than with oligoarthritis received DMARDs (OR=2.84, 95% CI 1.75 to 4.60, p<0.0001).

Conclusions Neither remission nor development of RA is delayed by GC treatment. Remission is rare in the first year of very early arthritis, occurring in <20% of the patients. Also, the need to start DMARDs was not influenced by GC treatment.

Statistics from


Arthritis is common in everyday practice and it is frequently unclear at the time of first occurrence if it will resolve without sequelae or become chronic. Rheumatoid arthritis (RA) is the most problematic chronic arthritis owing to its joint-destructive, often severely disabling character and the reduced life expectancy compared with that of the general population.1 In the very early stages RA cannot be easily differentiated from other types of early arthritis with a better prognosis, and diagnostic or classification criteria for early RA are still missing.2,,5 However, the approach to patients with inflammatory joint disease of short duration has changed over the past decade as it has become known that early therapeutic intervention is associated with better outcome.6,,11 To make early intervention possible, early referral to rheumatologists of patients with any type of inflammatory arthritis has been advocated in recent years and early arthritis clinics have been established12,,26 to allow for early recognition of RA and early institution of appropriate treatment. This early treatment may better prevent joint destruction and disability in comparison with delayed institution of disease-modifying antirheumatic drugs (DMARDs).3 8 27,,30 In addition, numerous studies have shown that in early RA, intensive treatment regimens, which mostly encompassed conventional DMARDs in combination with glucocorticoids (GCs) or biological agents, are better than less aggressive strategies.31,,34

On the other hand, many patients with early arthritis may experience spontaneous remission,2 35 36 may have diseases other than RA, may therefore not require DMARD treatment and thus will be overtreated when DMARDs are instituted. Therefore, one of the treatments advocated in these situations is an initial course of GCs. GCs are frequently used in general practice to relieve signs and symptoms of arthritis, but have also been shown to have disease-modifying effects.37,,39 However, the bulk of the literature on use of GC treatment refers to ‘established’ RA, using GCs singly or in combination with DMARDs.40, 41

In one of the few studies explicitly examining the problem of interference with very early arthritis, Green et al administered a single dose of a depot-GC (120 mg of methylprednisolone), either intra-articularly or intramuscularly. The results of this open study suggested that this approach was safe and efficacious.15 In a further study by the same group of investigators, Quinn et al42 demonstrated that, in a group of patients with early hand polyarthritis, application of 120 mg methylprednisolone was followed by disappearance of synovitis and/or remission in approximately 50% of the cases. If confirmed, this finding would indicate that very early treatment with a potent anti-inflammatory agent given for just short time may interfere with the evolution of the disease process, constituting an exceptional ‘window of opportunity’ to alter the fate of patients destined to develop chronic arthritis. Also, this frequently used approach would have an added value for general practice, beyond that of signs and symptom relief.

The aim of this trial was to investigate whether a single intramuscular dose of 120 mg methylprednisolone was more effective than placebo (PL) in inducing remission in patients with very early arthritis over a period of 12 months.

Study design, patients and methods

This was a double-blind, randomised, PL-controlled, international, multicentre, investigator-initiated clinical trial. A total of 29 sites participated: 26 in Europe, two in Mexico and one in Japan. Because patients with arthritis of short duration are frequently seen by rheumatologists after considerable delay owing to medical and non-medical factors,43 44 participating centres were selected on the basis of their ability to recruit patients with very early arthritis without preselection or uncontrollable bias. Thus, all the centres had either fast-track referral policies or ‘open-door clinics’ or both.


Patients with inflammatory arthritis of at least one joint and symptom duration of <16 weeks were included. A full list of exclusion criteria and details of the randomisation and blinding procedure can be found in online supplementary file 1.

Using the 66 joint count, patients were stratified according to the number of swollen joints into two groups: ‘monarthritis or oligoarthritis’ (≤3 swollen joints) or ‘polyarthritis’ (≥4 swollen joints).

The following assessments, as recommended by the European League Against Rheumatism (EULAR) and the American College of Rheumatology,45 46 were performed in every patient at baseline, after 2, 12 and 52 weeks: tender and swollen joints (66/68 joint count according to the EULAR handbook)47; patient-reported joint pain on a 10 cm visual analogue scale (VAS; 0=‘no pain’, 10=‘extreme pain’); patient- and evaluator-reported global disease activity on a 10 cm VAS (0=‘no activity’, 10=‘very high disease activity’); C reactive protein (CRP) and erythrocyte sedimentation rate.

Study treatment and concomitant medication

Consenting patients were given a single injection of 120 mg methylprednisolone (Depo-Medrol or Depo-Medrate) or an identical volume of PL (0.9% NaCl) intramuscularly.

Patients who did not use non-steroidal anti-inflammatory drugs (NSAIDs) at the time of study entry were prescribed paracetamol (to be taken as needed, up to 2 g/day). Patients who had already taken NSAIDs for their arthritis received celecoxib at 200–400 mg/day or continued NSAIDs to be taken as needed, plus a proton pump inhibitor, if indicated.

If initiation of treatment with DMARDs and/or additional (oral, parenteral or intra-articular) GCs was considered necessary by the doctors caring for the patient, the type of treatment and time point were recorded. In order to accommodate various treatment strategies in clinical practice in the different countries, choice and route of an additional GC or DMARD was not prespecified. When patients received additional GCs or DMARDs during the trial this was considered a treatment failure, although these patients were followed up until the end of 1 year after receiving the study drug.

End points

The primary end point was defined as the presence of clinical remission both at 12 and 52 weeks after the GC or PL injection, without further GC or DMARD treatment. Owing to lack of validated or published remission criteria for this patient population, remission was predefined by consensus of the steering committee members (KPM, RL, JSS, TWJH) as follows:

  1. No swollen joint and not more than two tender joints (unilateral MTPs (metatarsophalangeal joints) counting as one joint);

  2. Two of the three following criteria: (a) CRP within the normal range; (b) patient-reported global disease activity <1 cm on a 10 cm VAS; (c) pain <1 cm on a 10 cm VAS;

  3. No treatment (current or past) with DMARDs or GCs except for the study drug.

Secondary end points were the differences between treatment groups for the following items, among others:

  1. Fulfilment of the predefined remission criteria separately at weeks 2, 12 and 52;

  2. Time course of changes of the ‘core set variables’ for RA48 49 during the year of observation;

  3. The proportion of patients starting DMARDs during the first year;

  4. Final diagnosis at 52 weeks.

All data are presented as intention-to-treat analysis. When patients were lost to follow-up or received additional GCs and/or DMARDs during the time of follow-up, treatment was considered to have been a failure from that time point onward. The data of these patients were analysed using the last observation before DMARD/GC treatment or drop-out carried forward. For an additional sensitivity analysis, we assumed that all patients who were lost to follow-up and did not receive DMARDs up to their last visit went into remission (rather than assuming non-response as is usually done in clinical trial settings).

More details on financial support, design of the trial, ethical approval and preparation of the manuscript can be found in online supplementary file 2.

Statistical analyses

The power calculation, based on an expected spontaneous remission rate of 40% during 1 year of follow-up42 50 and a minimum clinically important difference of an additional 15% of remissions in the GC group called for inclusion of 400 patients (allowing for an estimated drop-out rate of 15%).

Statistical analyses were calculated using GraphPad Prism V 5.0 (GraphPad Software, 2007). Categorical variables were analysed using Fisher's exact or χ2 test, continuous variables by Mann–Whitney U or t tests, and results are expressed as p values and OR where appropriate. Analysis of variance (ANOVA) was used for exploratory secondary analyses comparing multiple groups. For all analyses, two-sided p values <0.05 were regarded as statistically significant. For multiple comparisons a Bonferroni correction was applied.


Patient disposition

A total of 389 patients were included (figure 1). Six patients were excluded from the analysis because of protocol violations (five because of prior use of GCs for their arthritis and one because symptom duration exceeded 16 weeks). Thus, 383 subjects were included in the intention-to-treat analysis. Their demographics are shown in table 1. One hundred and ninety-eight (51.7%) patients received methylprednisolone and 185 (48.3%) patients PL. Both treatment groups were similar with respect to age, gender distribution and duration of disease as well as the proportion of patients with monarthritis/oligoarthritis or polyarthritis (table 1). Of all 383 patients, 303 (79.1%) could be followed up for 12 months. Five patients attended only the first visit and did not return for follow-up. The proportion of subsequently lost patients was similar among the groups at week 12, and numerically (but not statistically significantly) higher in the GC-treated group than the PL group at week 52 (figure 1). Baseline characteristics of patients lost to follow-up—age, gender, duration of symptoms, clinical parameters, acute phase reactants and rheumatoid factor status—were similar to those of patients who completed the trial (data not shown).

Figure 1

Patient disposition. DMARD, disease-modifying antirheumatic drug; GC denotes glucocorticoid (methylprednisolone 120 mg) injection at first visit or oral (parenteral) GC subsequently; PL denotes placebo (saline) injection at first visit.

Table 1

Patient demographics at baseline

Effects of a single intramuscular injection of methylprednisolone as monotherapy on outcome in early arthritis

Primary end point

For long-term persistent remission, there was no difference between the GC- and PL-treated patients: only 32 (16.2%) of the 198 patients treated with GC achieved this primary end point (figure 2A); 33 (17.8%) of the 185 PL patients had persistent remission without an additional GC and/or DMARD (OR=1.13, 95% CI 0.66 to 1.92, p=0.68).

Figure 2

Remission (primary end point) was reached only in a minority of patients. (A) The primary end point was reached in 16.9% of all study patients. There was no difference between placebo (PL)- or glucocorticoid (GC)-injected patients. (B) The rates of remission in either group were identical at each visit. White bars, PL; black bars, GC.

In a sensitivity analysis assuming that all patients who were lost to follow-up and did not receive DMARDs up to their last available visit went into remission, remission rates were also not different between the two groups (32.4% for PL and 30.3% for GC, OR=1.10, 95% CI 0.72 to 1.70, p=0.66).

Secondary end points

DMARDs and additional GC treatment

The main reason for not achieving remission was persistent joint pain or swelling or raised CRP. Subsequent GC treatment (after the initial PL or GC injection at baseline) was needed in 98 patients: 48/198 (24.2%) patients in the GC group (11 patients without DMARDs) and 50/185 (27.0%) in the PL group (12 patients without DMARDs). In 86 GC patients (43.4%) and in 86 PL patients (46.5%) a DMARD was initiated (p=0.61).

Remission ‘on DMARD’ was seen in 39 (10.2%) patients (16 GC and 23 PL patients).

Remission frequencies at each follow-up visit

Fulfilment of remission criteria as defined for this protocol (see “Study design, patients and methods”) was analysed separately for each visit (figure 2B). As with the primary end point, there was no significant difference between PL- and GC-treated patients in the frequencies of remission at any of the studied time points.

Time course of changes of the core set variables

Analysis of the 28 joint count Disease Activity Score as well as individual core set variables is shown in figure 3. The data show that there was no difference between the groups at 3 months or 1 year. Only at week 2 patients receiving GCs had a significantly greater reduction in several of the variables than those receiving PL (p values given in figure 3 caption). However, this difference was not maintained over longer time periods. The transient 2-week response can be interpreted as an internal validation that this group of patients indeed received GCs.

Figure 3

Core set variables over time. Open squares, placebo; filled squares, glucocorticoid (GC). Values shown are medians and IQR except DAS28 (mean±SD). Patients receiving disease-modifying antirheumatic drugs (DMARDs) (and/or additional GC) were regarded as a failure of treatment for the primary end point. For these, the last observation without DMARD or additional GC was carried forward to week 52. There was a significant difference in favour of GC only at week 2 with regard to pain (p=0.0001), PGA (p=0.0008), EGA (p=0.0005) and DAS28 (p=0.0019), but not for the other variables. Subsequently no difference between the treatment groups could be discerned for any variable. CRP, C-reactive protein; DAS28, 28 joint count Disease Activity Score; EGA, evaluator-reported global disease activity; ESR, erythrocyte sedimentation rate; PGA, patient-reported global disease activity; SJC, swollen joint count; TJC, tender joint count; VAS, visual analogue scale.

Remission in patients initially presenting with oligoarthritis or polyarthritis

Among the 148 patients initially presenting with oligoarthritis, 41 (27.7%) achieved remission, whereas in the 235 patients with polyarthritis, 24 (10.2%) fulfilled the remission criteria as defined for this study (p<0.0001). There was also a statistically significant difference in the duration of symptoms between patients with oligoarthritis and those with polyarthritis: patients with oligoarthritis were seen after a median (IQR) of 45 (17, 82) days from symptom onset, whereas patients with polyarthritis had their baseline visit 60 (32, 89) days after their (self-reported) first symptoms (p=0.0020, Mann–Whitney test).

Within both strata, again, there was no significant difference in remission rates between PL- and GC-treated patients (data not shown).

Remission in patients with different symptom duration

In an exploratory post hoc analysis, frequencies of remission were analysed in relation to duration of symptoms. Patients were grouped by ‘weeks after symptom onset’ (0–4, 5–8, 9–12 and 13–16 weeks, respectively), and remission rates between the groups were compared by ANOVA. There was a sharp and highly significant decline in remission rates after week 8 in both the patients with oligoarthritis and those with polyarthritis (figure 4). Whereas in patients who were seen within 4 weeks from symptom onset, almost 45% of patients with oligoarthritis attained remission (and 22% of those with polyarthritis), patients presenting with >8 weeks of symptom duration had a chance of achieving persistent remission of only about 10% (p<0.0001). Also in this exploratory analysis, no difference between treatment with GC compared with PL could be discerned. Thus, the cut-off point for a high likelihood of disease persistence appears to be around 8 weeks.

Figure 4

Exploratory post hoc analysis by time after symptom onset. Patients presenting earlier were highly significantly more likely to achieve lasting remission in both the oligoarthritis and polyarthritis group (p=0.0034 for oligoarthritis (white bars), p=0.0051 for polyarthritis (black bars), p<0.0001 for all patients (grey bars)). n denotes number of patients each month. In the table, absolute numbers of patients achieving remission/total patients in each respective group are shown (all comparisons p>0.05 by Fisher's exact test, except *difference between remission in oligoarthritis versus polyarthritis week 5–8: p=0.0168).


On the 303 patients followed up for 1 year, the final diagnosis assigned by the investigators was documented in the case report forms. Almost 50% of the patients were diagnosed as RA. In line with previous observations,51 52 approximately 30% of the patients were labelled as ‘undifferentiated arthritis’ at the 52-week visit or maintained this labelling from baseline (table 2). A significantly larger percentage of patients who presented with polyarthritis were subsequently assigned a diagnosis of RA (113 of 192 patients, 58.9%) compared with patients with oligoarthritis (32 of 111, 28.8%, p<0.0001). The proportions of patients developing RA were similar in the GC and PL groups (45.1% vs 50.7%, respectively, p=0.36). Thus, GC could not reduce or prevent the evolution of RA.

Table 2

Diagnoses in patients followed up for 12 months (numbers of patients)

Adverse events

During the 1-year observation period, 22 serious adverse events occurred in 20 patients (none of them classified as serious unexpected adverse drug reactions): in one patient treated with methylprednisolone worsening polyarthritis necessitated hospital admission; later, the same patient developed oedema of the lower extremity suspected to be deep vein thrombosis and leading to hospital admission (deep vein thrombosis was not confirmed). In another patient (treated with PL) one admission due to severe hypertension and a subsequent admission for cataract surgery were necessary. The remaining 18 adverse events occurred in 12 patients treated with methylprednisolone and in six PL patients (p=0.2316). Only one of these adverse events was classified by the treating doctors as related to the study drug; it constituted worsening of pustulous psoriasis, which occurred 10 days after injection of methylprednisolone and resolved without sequelae.


In this trial we did not find any evidence for a long-lasting disease-modifying effect of a single dose of 120 mg methylprednisolone intramuscularly, such as an increase in the rate of remissions, in patients with very early arthritis (up to 16 weeks of symptom duration). Because guidance of clinical practice by trial results is usually hampered by the fact that inclusion criteria often do not match the patients presenting in clinics or offices,44 53 the trial was designed to be ‘real-life-like’ as possible. Also, trials focusing at daily clinical practice are generally performed in one country or one centre, such as strategic trials.34 54 Therefore, we performed a trial that was designed to generate representative data for all patients with early arthritis world wide in a double-blind, PL-controlled design. The findings indicate that, while the intramuscular GC provided rapid relief of signs and symptoms in recent-onset inflammatory arthritis, this effect lasted only for a short time.

Only about 17% of the patients, whether treated with GC or PL, achieved remission, which was defined by stringent criteria such as the absence of any joint swelling and normal CRP without additional DMARD or GC treatment. In the majority of the patients, arthritis signs and symptoms were not normalised. Further, the proportions of patients needing DMARDs did not differ between the groups. The only significant effect attributable to GC treatment was a reduction in pain and global disease activity after 2 weeks. This is in line with the known anti-inflammatory effect of GCs. However, this difference was not maintained.

Exploratory post hoc analysis showed that patients first seen >8 weeks after symptom onset had a chance of only about 10% of achieving sustained remission, while this rate was significantly higher in patients with <8 weeks of symptoms, probably owing to spontaneous remissions. This suggests that recent-onset arthritis will become persistent in 90% or more of subjects whose joint symptoms extend beyond 8 weeks, while there may be a greater tendency towards spontaneous remission for a significant proportion of patients before that time.

Our study has several limitations. The primary end point requiring persistent remission starting at 3 months from randomisation may have been too stringent. However, even at the 12-month end point the frequency of remissions was similar to that of the combined 3- and 12-month time points and there was no significant difference between the treatment groups at any of the prespecified time points. Another limitation relates to the fact that we assumed that patients who were lost to follow-up were non-responders, as is usually done in clinical trials. Although it might well be that these patients did not return because their symptoms had subsided, the sensitivity analysis in which we assumed all patients lost to follow-up but not treated with DMARDs to be responders, did not show a difference in the primary end point between GC and PL. Even assuming this ‘optimal outcome’ scenario in this analysis, remission would have been observed in only 30% of the patients. A third limitation relates to possible inclusion bias: although all participating centres had either fast-track referral policies or ‘open-door clinics’ or both, the possibility cannot be excluded that some patient candidates may have been reluctant to enter a PL-controlled trial at the very beginning of a possibly disabling disease. Nevertheless, this would probably not have influenced the main results of this trial significantly.

The data obtained are remarkable in several respects. In contrast to some observations,42 50 the rate of spontaneous remission in patients with very early arthritis is low, since in more than 80 of 100 patients with early arthritis, disease persists. This is in line with the recently published results of the STIVEA trial and confirms the findings of this British study55 in an international double-blind randomised controlled trial. Another recently published small PL-controlled study on GCs in patients with arthralgia (not arthritis)56 also did not find any effect of the intervention on subsequent frequency of arthritis. GCs, which have highly potent anti-inflammatory and immunomodulating activity, can apparently not interfere with the progress of the disease, even in the majority of patients whose symptoms have been present for <8 weeks. GCs do not reduce the need to start DMARD treatment, and the GC injection was not effective in delaying the need to start DMARD treatment for a significant period of time. Lastly, spontaneous remission rates drop sharply, and earlier than previously believed—namely, after only 8 weeks of symptom duration.

In conclusion, <20% of the patients with early arthritis attain spontaneous remission, and this is less frequent for those with a polyarticular disease than for those with oligoarticular arthritis at the start. Consequently, treatment with traditional DMARDs was needed in most patients. It is evident, therefore, that even in this early arthritis population the majority of patients will require more intensive treatment. Thus, early arthritis is anything but a benign disease and a therapeutic approach which employs only GCs and expectant observation of the patients will be insufficient in the vast majority of people with arthritis.


The authors owe special thanks to Love Amoyo, Michaela Ernst, Lorenz Machold, Mona Mathis, Birgit Prodinger, Enijad Sahinbegovic and Carl-Walter Steiner for data management assistance.


View Abstract


  • Funding European League Against Rheumatism, Pfizer, AutoCure.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the ethics committee of the Medical University of Vienna and ethics committees of all participating centres.

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

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.

Linked Articles

  • Miscellaneous
    BMJ Publishing Group Ltd and European League Against Rheumatism