Doi:10.1016/s0140-6736(97)01300-7

Randomised comparison of combined step-down prednisolone,methotrexate and sulphasalazine with sulphasalazine alone in earlyrheumatoid arthritis Maarten Boers, Arco C Verhoeven, Harry M Markusse, Mart A F J van de Laar, René Westhovens, J Christiaan van Denderen, Derkjen van Zeben, Ben A C Dijkmans, André J Peeters, Piet Jacobs, Hans R van den Brink, Hubert J A Schouten, Désirée M F M van der Heijde, Annelies Boonen, Sjef van der Linden suppressed damage progression, whereas sulphasalazine did so less effectively and with a lag of 6 to 12 months. There Background The value of intensive combination therapy in were fewer withdrawals in the combined therapy than the early rheumatoid arthritis is unproven. In a multicentre, sulphasalazine group (6 [8%] vs 23 [29%]), and they double-blind, randomised trial (COBRA), we compared the combination of sulphasalazine (2 g/day), methotrexate (7·5 mg/week), and prednisolone (initially 60 mg/day, tapered in additional disease control over and above that of 6 weekly steps to 7·5 mg/day) with sulphasalazine alone. sulphasalazine alone that persists for up to a year after Methods 155 patients with early rheumatoid arthritis corticosteroids are stopped. Although confirmatory studies (median duration 4 months) were randomly assigned and long-term follow-up are needed, this approach may prove combined treatment (76) or sulphasalazine alone (79).
useful in the treatment of early rheumatoid arthritis.
Prednisolone and methotrexate were tapered and stopped after 28 and 40 weeks, respectively. The main outcomes were the pooled index (a weighted change score of five disease activity measures) and the Sharp/Van der Heijde radiographic damage score in hands and feet. Independent health-care professionals assessed the main outcomes The treatment of rheumatoid arthritis is traditionally without knowledge of treatment allocation.
characterised by escalation. The first step is non-steroidal Findings At week 28, the mean pooled index was 1·4 (95% anti-inflammatory drugs (NSAIDs), and then if necessary CI 1·2–1·6) in the combined treatment group and 0·8 a sequence of progressively toxic second-line drugs(disease-modifying antirheumatic drugs) is introduced.1 (0·6–1·0) in the sulphasalazine group (p<0·0001). At this There is evidence that some of these disease-modifying time, 55 (72%) and 39 (49%) patients, respectively, were drugs provide a degree of disease control2—ie, decrease improved according to American College of Rheumatology disease activity but also maintain or improve physical criteria. The clinical difference between the groups function and retard radiographic joint damage.3 However, decreased and was no longer significant after prednisolone both patients and physicians are dissatisfied with the long- was stopped, and there were no further changes after term results of traditional therapy. A 1996 study suggested methotrexate was stopped. At 28 weeks, the radiographic that early introduction of disease-modifying antirheumatic damage score had increased by a median of 1 (range 0–28) drugs may be more beneficial than delayed introduction for in the combined-therapy group and 4 (0–44) in the patients with recently diagnosed rheumatoid arthritis.4 sulphasalazine group (p<0·0001). The increases at week 56 Research is focused towards finding new, more effective (2 [0–43] vs 6 [0–54], p=0·004), and at week 80 (4 [0–80] drugs, reassessment and earlier use of existing drugs (such vs 12 [0–72], p=0·01) were also significant. Further analysis suggests that combined therapy immediately The COBRA trial (Combinatietherapie Bij Reumatoide Artritis) is an adaptation of the latter two options—step- University Hospital, Maastricht (M Boers MD, A C Verhoeven MD, down bridge therapy with corticosteroids in early H J A Schouten, D M F M van der Heijde, A Boonen, rheumatoid arthritis.1 Our intention was to control disease S van der Linden); Zuiderziekenhuis, Rotterdam rapidly at a very early stage, with agents that have (H M Markusse MD); Medisch Spectrum Twente and Twenteborg overlapping windows of efficacy onset; and then, after 6 Hospital, Enschede and Almelo, Netherlands months to taper and stop the more toxic components while (M A F J van der Laar MD); Pellenberg Hospital, Catholic University, retaining disease control. We devised a regimen comprising Louvain, Belgium (R Westhovens MD); Jan van Breemen Instituut, a short period of high-dose oral prednisolone, rapidly Amsterdam (J C van Denderen MD); Bronovo Hospital, The Hague tapered to a low maintenance dose. As the other (D van Zeben MD); University Hospital, Leiden (B A C Dijkmans MD); components we chose methotrexate, commonly used as the Reinier de Graaf Gasthuis, Delft (A J Peeters MD); Sint LaurentiusHospital, Roermond (P Jacobs MD); Medisch Centrum Alkmaar, disease-modifying drug of first choice in the USA, and sulphasalazine as the anchor drug to remain after the othertwo drugs were withdrawn. In Europe, sulphasalazine is Correspondence to: Dr Maar ten Boers, Depar tment of ClinicalEpidemiology, Free University Hospital, PO Box 7057, 1007 MB commonly used as the disease-modifying drug of first We carried out a 56-week multicentre, randomised tablets of 500 mg, Pharmacia & Upjohn, Uppsala, Sweden) 500 controlled trial among patients with early, mg/day, increased to 2000 mg/day over a period of 3 weeks. In rheumatoid arthritis to study the degree of disease control addition, the combination therapy group received prednisolone afforded by a combination of sulphasalazine, methotrexate, and methotrexate; the control group received matching placebotablets and capsules identical in appearance and taste. The daily and high/low oral prednisolone given in the first 28 weeks, prednisolone dose was 60 mg in week 1, 40 mg in week 2, 25 mg compared with that achieved with sulphasalazine alone; in week 3, 20 mg in week 4, 15 mg in week 5, 10 mg in week 6, and to find out whether control could be maintained on and 7·5 mg thereafter (week 1–6—one gelatine capsule containing the daily dose, capsule compound by Bufa, Uitgeest, Netherlands; withdrawal of prednisolone and methotrexate in the second week 7 and later—5 mg tablets by CentraFarm Nederland bv, Etten-Leur, Netherlands; some of these tablets were broken by thepharmacy so that 7·5 mg could be taken daily). The cumulative dose over the first 6 weeks was 1190 mg; over the first 28 weeks itwas 2345 mg, corresponding to a mean of 12 mg daily. The methotrexate prescription was 7·5 mg in a single weekly dose We recruited patients between May, 1993, and May, 1995, in ten (PharmaChemie bv, Haarlem, Netherlands). If an adverse event centres (nine in the Netherlands, one in Belgium). To optimise the occurred, the drugs were temporarily withdrawn, and reintroduced benefit/risk ratio in line with the study purpose, we applied strict at lower doses according to a fixed protocol where possible.
eligibility criteria to include patients with early rheumatoid arthritis Prednisolone and methotrexate were stopped after 28 weeks and who had very active disease and were most likely to benefit from 40 weeks, respectively. Both drugs were gradually withdrawn to this intensive treatment, in whom effects could be easily measured, decrease the chance of a disease flare. Thus, from week 29 to 35, a and in whom we believed adverse effects would be least likely. The day of zero prednisolone dose was introduced each week: first inclusion criteria were: a diagnosis of rheumatoid arthritis (American College of Rheumatology criteria7) with onset of disease at or after 16 years of age; active disease of the joints and prednisolone on Monday, Wednesday, and Friday; until after 6 inadequate control of arthritis (due to lack of efficacy or toxicity of weeks, the prednisolone had been stopped. After 40 weeks of treatment); and treatment with NSAIDs in adequate doses for at treatment, methotrexate was tapered: the drug was given at 5 mg least 3 months. Such treatment could already have been initiated at per week for 3 weeks, then at 2·5 mg per week for 3 weeks, then the start of symptoms, not necessarily at the time of diagnosis. We stopped. If there was a flare in disease activity, the last medication defined disease activity as the presence of six or more actively stopped was reintroduced. A flare was defined per protocol as an inflamed joints, located at three or more different sites (a site is increase of five in active joint count or an increase from zero to defined as either one large joint or a group of small joints: the three compared with the situation at week 28 (an active joint is joints of the wrist, the metacarpophalangeals, the proximal hand swollen or tender on pressure; counting of joint groups in one interphalangeals, the distal hand interphalangeals, ankles, the hand or foot as above). If the research medication had to be tarsometatarsals, the metatarsophalangeals, and the proximal and stopped for any reason and a consecutive disease-modifying distal foot interphalangeals) and presence of at least two of the antirheumatic drug was started, the protocol recommended that a following: nine or more tender joints (irrespective of site), morning drug not in the combination should be given, preferably stiffness of 45 min or longer, and a Westergren’s erythrocyte intramuscular gold salts. After 56 weeks, the protocol ended, and sedimentation rate (ESR) of 28 mm or more in the first hour.
the treating physician was at liberty to change second-line therapy, We excluded patients who had had rheumatoid arthritis for or to attempt a second tapering of methotrexate or prednisolone in longer than 2 years, those previously or currently treated with any those patients still on combination therapy. Where possible, disease-modifying antirheumatic drug except antimalarials (eg, blinded protocol treatment was continued. To maintain allocation gold, d-penicillamine, azathioprine, or cyclophosphamide) or concealment for other patients still in the protocol, the treatment corticosteroids (for arthritis or another disease), code was revealed only for those patients still on combination comorbidity or recent (within the 3 months before enrolment) major surgery, or inability to comply with the protocol. Adequatecontraception was required. Further exclusion criteria were age below 18 or over 70; hypersensitivity to study medication, sulpha- NSAIDs and simple analgesics were allowed; discontinuation was containing compounds, or aspirin; hypersensitivity to three or actively pursued. A maximum of two intra-articular steroid more drugs; active infectious disease; a history of tuberculosis, injections was allowed in two periods after week 38 of the protocol, but not in the 6-week period preceding independent assessment.
gastrointestinal ulceration; any history of gastrointestinal bleeding Any other intervention with parenteral or oral corticosteroids was or neoplasia; diabetes mellitus; hypertension treated with more not permitted. All patients received folic acid (1 mg/day) during than one antihypertensive drug; significant cardiovascular disease; methotrexate or placebo prescription. Vitamin D deficiency liver disease; cataract; glaucoma; haematological disorders; partial apparent at the laboratory screening before inclusion was or total colectomy; reduced renal function (creatinine clearance <50 mL/h) proteinuria (>0·5 g/day); hypoalbuminaemia; chronicdermatitis; treatment with phenytoin, phenylbutazone, salicylates, Patients who met the eligibility criteria were entered into the study (dicoumarol derivatives); and a history of alcohol or substance and assigned a unique study identification number by telephone.
abuse (ie, inability to limit alcohol intake to a maximum of 70 g This number implied random allocation to one of the two weekly) or use of any experimental drug less than 2 months before treatments with stratification by centre. For each centre, a separate randomisation list was generated by an adaptive biased urn The study protocol was approved by research and medical algorithm. In contrast to fixed blocks, this algorithm ensured that ethics committees in all participating hospitals. The patients were the rheumatologists would have no clue to the allocation of each fully informed about the potential side-effects of all the drugs. To subsequent patient in a setting where unblinding was possible; yet maintain allocation concealment, they were told that response to it also guaranteed an approximately equal distribution over the treatment was variable in onset and efficacy with all three drugs.
groups even in the centres with smaller numbers of patients.8 The All patients gave written informed consent.
assignment was known only by the employees of the MaastrichtHospital pharmacy who prepackaged the medication; it was disclosed to the treating physician only in case of an emergency.
Both groups received sulphasalazine (Salazopyrine enteric-coated Primary analysis was done with coded group allocation after entry of all study data. The full randomisation codes remained concealed reactant (ESR) included in the pooled index, these measures are until completion of the primary analysis.
swollen joint count (48 joints: modified from American College ofRheumatology 66-joint count;11 small foot joints as one joint site and no midfoot joints), pain (assessed by the patient on a 100 mmvisual analogue scale; worst imaginable pain at the right anchor), Each centre was staffed by a rheumatologist, a research nurse, and and patient’s overall assessment (on a 100 mm visual analogue an independent assessor. The rheumatologist was responsible for scale, worst and best imaginable health status at the left and right the identification and inclusion of the patients, and for all medical anchor, respectively).To facilitate comparisons with other studies, policy decisions. The research nurse monitored safety through regular follow-up schedules (first weekly, then every 4 weeks) and questionnaire was complemented with the more generic health also measured disease activity. Independent assessors (mostly assessment questionnaire (Dutch validated version; scores 0–3, 3 physiotherapists) applied the outcome measures at baseline and at weeks 16, 28, 40, and 56; in almost every instance a patient was We expressed improvement in individual patients by the seen by the same assessor. These health professionals were not American College of Rheumatology preliminary criteria for involved in care of patients; they were also asked not to discussdisease activity or the treatment with the patients. Independent remission16 (occurrence and duration; because no inquiry on assessment ensured optimum concealment of primary outcome fatigue was made we used the concept of a “probable remission’ assessments, especially important in the first 6 weeks of the for instances in which a patient would be in remission when protocol, when potential effects and side-effects of high-dose absence of fatigue was assumed). Furthermore, we used the prednisolone would be most apparent. These assessments included American College of Rheumatology preliminary criteria for all primary and core-set outcome measures except pain score, grip strength, and ESR. The follow-up schedule is continuing; all improvement in tender and swollen joint counts plus a similar outcomes during the first 56 weeks are reported here. In response improvement in at least three of five remaining core-set measures).
to criticism about the follow-up period for the radiographs, we Before calculating improvement percentages, we ensured (by read and analysed the 80-week radiographs at a later stage.
recoding if necessary) that all scales decreased on improvement.
Before the study and then once a year, all study personnel We also report improvement with application of a 50% threshold trained together to maintain assessment quality and agreement instead of the 20% in the American College of Rheumatology between observers. A specially designed manual of procedures and assessment techniques was available in each study centre. The trial To facilitate comparison with other European studies, we report was coordinated and data managed in the University Hospital Maastricht. Safety and toxicity were monitored by a safety containing the Ritchie tender joint index (RI), swollen joint count committee of two independent rheumatologists, the Maastricht (JC), ESR, and patient’s overall assessment University pharmacist, and a statistician (HJAS). The pharmacy Score=0·54͌RI+0·065JC+0·331nESR+0·007OA centre of the University Hospital Maastricht was responsible for The term disease-controlling has been suggested for drug production, packaging, and distribution.
antirheumatic treatment regimens that improve disease activity,retain or improve physical function, and decrease progression ofradiographic damage.3 A priori, we expected out study to be too small to detect small differences in radiographic progression The primary endpoint of the therapeutic intervention was a pooled between the two groups, since both were treated with the disease- index summarising the change in five measures after 28 weeks of treatment. Pooling is a validated method to increase sensitivity of observers (AB and ACV) assessed radiographic damage, unaware separate measures.9 To obtain the pooled index of one of the of the identity of the patients. They separately scored radiographs groups at week 28, we calculated a standardised change score of of hands and feet according to van der Heijde’s modification of that group by dividing the mean change in one measure by its Sharp’s method.2 This method reflects erosions and joint-space pooled SD of change at week 28. This procedure was repeated for narrowing in 44 joints in the hands and feet. The principal each of the five measures; the pooled index is the mean of the measure, the total score, is the sum of erosion and narrowing standardised scores. To obtain pooled index values for another scores, and ranges from 0 to 448. The method requires time point, change scores at that point were standardised through radiographs to be presented in ordered fashion (baseline, and division by the same pooled SD at week 28. Finally, a constant was weeks 28, 56, and 80). Scores can either be stable or increase; added to all index values so that the value at baseline was zero. We decrease (indicating improvement) is not possible. We report the selected five measures for maximum sensitivity to change:10 Tender mean of the two observers’ erosion, narrowing, and total scores.
joint count (68 joints11); overall assessment by the independent As an exploratory analysis, we also report the cumulative assessor (on a 100 mm visual analogue scale, worst and best number of joints free of erosions at baseline in which at least one imaginable health status at the left and right anchor, respectively); erosion developed during follow-up. For this purpose, joints were grip strength (by vigorimetry; Martin, Tottlingen, Germany, range grouped into four areas on each side: wrist (six joints), 0–150 kPa, mean of medians of three measurements in both metacarpophalangeal (five joints); proximal interphalangeal (four hands);12 ESR (Westergren method); and McMaster Toronto joints); and foot joints (six joints). The first erosion in each area arthritis questionnaire,13 which follows improvement in five was counted. Furthermore, we explored the rate of radiographic impaired activities elicited and priority-ranked by the patient change in each of the measurement periods by calculating not only during a baseline interview, together with change scores for quality the change scores from baseline, but also the change scores of life, psychological, social, and emotional wellbeing. The scores of between week 28 and 56 and between week 56 and 80.
this questionnaire reflect change, increase as disability improves,and vary from 10 (maximum deterioration) to 40 (maximumimprovement). In its original format, the baseline questionnaire score differs from the follow-up scores because the change items Laboratory monitoring at every control visit comprised complete are not available. To make these scores directly comparable, we added mock change items at baseline and scored them as aminotransferases, alkaline phosphatase, creatinine, blood urea “unchanged”. Grip strength and ESR were assessed by the nitrogen, and electrolytes, and urinalysis for glucose and albumin.
research nurses every week at the start of the protocol, then at least Toxicity was assessed by counting of each adverse event reported and possible subsequent changes to treatment (eg, withdrawal).
We assessed all remaining disease activity measures of the World Each patient underwent pulmonary function tests (expiratory Health Organisation/International League of Associations for volume and carbon monoxide diffusion capacity) at baseline and Rheumatology core set as secondary endpoints.14 As well was twice yearly thereafter. Bone densitometry was done by an tender joint count, assessor’s overall assessment, and acute phase operator unaware of treatment assignment, in all centres where a Figure 1: Trial profile*Patient lost to follow-up.
AE=adverse events; LE=loss of efficacy; other includes protocol violations.
dual-energy X-ray absorptiometer was available (Lunar, Hologic, the pooled index (SD 0·45) between the treatment groups at two- or Norland, in eight centres). We report changes in bone mass for sided ␣=0·05, given a maximum dropout rate of 50%. Enrolment lumbar spine and femoral neck (mean of right and left hip).
stopped at 156 patients for practical reasons, because the actual We assessed IgM rheumatoid factor serostatus in a time- dropout rate was 20%. All analyses were based on intention to resolved fluoroimmunoassay (rabbit IgC antigen; Nordic, Tilburg, treat as initially assigned. All available data were used. Data Netherlands); values over 20 kU/L classified patients as positive for missing due to loss to follow-up were handled by a last observation rheumatoid factor.19 Class II HLA genotype was identified by carried forward approach. For other missing data, values were serological typing (Tissue Typing Laboratory, interpolated if actual assessments were available at least every 28 weeks. No interim analyses were done.
Measures with a Gaussian distribution are expressed at baseline as mean and SD, and as mean change with 95% CI. The mainendpoint was initially analysed by two-way ANOVA with We assessed compliance by tablet counts at every control visit, by treatment, centre, and their interactions as factors. The latter two questioning (including a quantification of the number of tablets factors were not significant (centre p=0·07; interaction p=0·79). In missed), and by measurement of sulphapyridine (a sulphasalazine view of the large effect of treatment, further analyses ignored metabolite) in urine samples taken at weeks 16, 28, and 40. We centre as a factor, with exception of the multivariate analysis.
classified as non-compliant all patients on protocol treatment who Measures with a non-Gaussian distribution are expressed as were negative for sulphapyridine once or who failed to return median and median change (range) and analysed with Mann- tablet boxes at control visits more than once. In the first 28 weeks, Whitney tests; measures with a discrete distribution are expressed we made judgments at every control visit, and classified as as counts (%) and analysed by continuity-corrected ␹2 tests or “probably non-compliant” patients who in the first 28 weeks on Fisher’s exact tests where appropriate. The level of significance was average missed more than one daily dose per week of set at p<0·05, two-sided. No adjustment was made for multiple sulphasalazine or prednisolone, or more than one weekly dose of methotrexate over a period of 6 weeks.
For the main clinical and radiological outcome, multivariate analyses tested whether imbalance in important prognostic factors between the two groups despite randomisation affected the study The target sample size was 168 patients. This number yields a results. The dependent variables were the pooled index and the power of at least 90% to detect a difference of 0·33 or greater in progression in total radiological damage score at 28 weeks. The latter was log-transformed (new variable=log[total change score +1]) because of a skewed distribution. Full and parsimonious models were constructed: the full models contained the predictive variables: treatment group, centre, sex, age, disease duration, rheumatoid factor status, presence of HLA genotype DR4 or DR2, number of years education, marital status, and baseline questionnaire, and radiological damage. The parsimonious models contained only those variables selected in forward stepwise duration in monthsPrevious treatment with regression analysis (F to enter >4·0). Plots of residuals versus fitted and versus predictor values were inspected for departure of regression analysis assumptions. Because there was still substantial skewness after log transformation on the total radiological damage score, the results of the regression analysis on this variable must be *Homozygotic or heterozygotic DR type, assessed in 143 patients (92%).
The trial included 156 patients (figure 1). In one patient †Patients with available baseline radiographs; combined treatment group n=74,sulphasalazine group n=75.
the protocol medication was stopped within 1 weekbecause his disease was in spontaneous remission at Table 1: Baseline characteristics of study patients baseline. Data for this patient are not reported. 76 patientsreceived combination treatment, 79 sulphasalazine only.
remissions (32%) in the combined-treatment group, Five patients (3%, all in the sulphasalazine group) were lost compared with 14 and five in the sulphasalazine group to follow-up before week 56. In six patients (all withdrawn) (24%; p=0·38). Of these, only one patient in the the treatment assignment had to be revealed before week combined-treatment group and three in the sulphasalazine group had persisting remission at 56 weeks.
The two treatment groups were similar in terms of The difference in clinical efficacy between the treatment groups decreased and was no longer significant after the demographic and other prognostic variables (table 1).
withdrawal of prednisolone, and there were no further Except for one Asian patient in each group, all patients changes when methotrexate was withdrawn (figure 2).
were white. Our eligibility criteria resulted in a study group However, differences in two of three measures of physical with very early, active, and severe rheumatoid arthritis; in function (health assessment questionnaire, grip strength) 77% the trial medication was the first disease-modifying remained near significance (table 2). Prednisolone was antirheumatic drug. High rates of rheumatoid-factor restarted (for disease flares) in six patients in the combined-treatment group. Methotrexate was restarted in radiographic damage all indicate a poor a-priori prognosis 13 patients in the combined-treatment group, and (table 1). In 21% of all patients (ie, 32% of those with baseline erosive disease) erosions were initially found only Adjustment for prognostic variables did not change the difference in efficacy between treatments (crude coefficient for the additional effect of combined treatment vs Within a few weeks, combined therapy greatly improved sulphasalazine on the pooled index at week 28: 0·63 [SE disease activity in most patients (figure 2, table 2).
Sulphasalazine also improved disease activity substantially, although less than combined treatment. There was an physical disability score and disease duration, as well as almost immediate response to combined treatment in all study centre, significantly affected the pooled index at week frequently assessed measures (eg, grip strength, pain, ESR; 28; patients with high initial disability and shorter disease the latter is shown in figure 2). Despite a daily duration were more likely to improve. The effect of prednisolone dose of only 7·5 mg from week 7 onwards, treatment was not changed in the model incorporating further improvement continued up to week 28. At this time, the clinical efficacy of combined treatment wasalmost double that of sulphasalazine (figure 2, table 2). 55 (72%) of the combined-treatment group compared with 39 Radiographs of 147 patients (73 combined treatment, 74 (49%) of the sulphasalazine group had improved according sulphasalazine) were available for assessment (baseline and to “20%” American College of Rheumatology criteria at least one during follow-up). The between-observer (p=0·006), and 37 (49%) compared with 21 (27%) had reliability of the assigned total scores was satisfactory improved under the “50%” criteria (p=0·007).
(within-class correlation coefficient of absolute scores 0·91, In the combined-treatment group 16 patients had and of change scores 0·88). Because of skewness in the probable remissions and five definite remissions during the data we also calculated Spearman rank correlations first 28 weeks (total 28%). In the sulphasalazine group, the (absolute scores 0·90, change scores 0·89).
corresponding numbers were nine and four (total 16%; The groups were well balanced in terms of damage at baseline (figure 3, table 3). 26% versus 22% of patients had remissions clustered near the beginning and end of the first no erosions; 23% versus 19% had a total score of 0. The 28-week period. Almost all of these remissions ended when total score had increased significantly more in the prednisolone was stopped, and in the second 28 weeks, sulphasalazine group than in the combined treatment only a few additional patients had remissions. For the total group at 28 weeks (p<0·0001), 56 weeks (p=0·004), and study period, there were 18 probable and six definite 80 weeks (p=0·01; table 3, figure 3). The differences Figure 2: Clinical outcomes of treatment, expressed as mean (95% CI) pooled index and changes in its component partsPositive values indicate improvement in pooled index, grip strength, and MACTAR (McMaster Toronto ar thritis) questionnaire. Negative valuesindicate improvement in the remaining measures. Changes in ESR in first 16 weeks are shown in graph in upper right corner (note different timescale).
between the groups were greatest for increases in the difference in efficacy between treatments (crude coefficient erosion score (table 3). At 28 weeks, 31% of the combined- for the additional effect of combined treatment vs treatment group versus 13% of the sulphasalazine group sulphasalazine on the log-transformed total score at week had stable scores (p=0·009); at 56 weeks, 19% versus 10% 28 Ϫ0·302 [SE 0·075], p<0·0001; adjusted coefficient (p=0·23); at 80 weeks, 10% versus 7% (p=0·75). In both Ϫ0·297 [0·062], p<0·0001). In stepwise regression groups, patients without erosions at baseline showed little analysis, baseline values for radiological damage, disease progression over 80 weeks: median increase 2 (0–25).
activity score, HLA-DR4 genotype, and rheumatoid factor The findings on new erosive damage appearing in one of as well as centre significantly affected the progression of eight areas previously free of erosions were similar. After 28 radiological damage at week 28. Patients who had higher weeks such damage had appeared in median zero areas baseline values for radiological damage and disease activity (0–5) in the combined-treatment group, compared with score and who were positive for HLA DR4 and rheumatoid one (0–6) in the sulphasalazine group (p<0·0001). The factor, had higher rates of progression. Again, the effect of corresponding numbers at 56 weeks were zero (0–6) and treatment was not changed in the model incorporating one (0–7, p<0·0001), and at 80 weeks one (0–7) and two The damage rates calculated per study period showed slow progression of damage in the combined-treatment Significantly fewer patients stopped combined treatment group in the first period (baseline to 28 weeks), whereas than stopped sulphasalazine (table 4, p=0·0008), and progression in the sulphasalazine group was more rapid combination patients dropped out later. Differences were (median increase in total score 0–28 weeks 1 [0–28] in apparent for both toxic effects and lack of efficacy. For combined-therapy group vs 4 [0–44] in sulphasalazine instance, all four withdrawals from the combined-treatment group; p<0·0001). In the second period, in which group because of lack of efficacy occurred after week 28, prednisolone and methotrexate were stopped (28–56 when prednisolone and methotrexate were stopped; by weeks), the rate of progression was again lower in the contrast, most of the 19 withdrawals for lack of efficacy in combined-treatment group than the sulphasalazine group the sulphasalazine group occurred before week 28 (figure (median increase 1 [0–36] vs 2·5 [0–27], p=0·04), but 1). The adverse reactions that led to withdrawal of two during weeks 56–80 the rate of progression did not differ significantly (1·5 [0–36] vs 2·5 [0–32], p=0·37).
gastrointestinal-tract complaints and dyspnoea (final Adjustment for prognostic variables did not change the diagnosis exacerbation of chronic bronchitis). In the SSZ=sulphasalazine alone; VAS=visual analogue scale; MACTAR=McMaster Toronto arthritis questionnaire.
*Combined minus sulphasalazine; positive values indicate better average outcome in combined treatment group.
†Positive values for change indicate improvement, Table 2: Main clinical outcomes of treatment after 28 and 56 weeks of combined (n=76) and sulphasalazine (n=79) treatment sulphasalazine group the adverse events leading to yearly bone densitometry where possible (64 combined withdrawal were rashes in four patients, gastrointestinal- treatment, 62 sulphasalazine patients assessed). Patients in tract complaints in two (one with concurrent proteinuria), both the combined-treatment and sulphasalazine groups gained weight (mean gain at 28 weeks 2·5 kg [95% CI 1·8, aminotransferases in one patient, and thrombocytopenia 3·2] vs 0·7 kg [Ϫ0·2, 2·2], p=0·002; at 56 weeks 1·7 [0·8, (diagnosis preleukaemic disease) in one.
2·6] vs 1·2 [0·2, 2·2] kg, p=0·49). Blood pressure remained The study medication was discontinued and restarted at an adjusted dose according to protocol in five patients.
prognosis factors for osteoporosis were balanced and mean Three of these patients (one in the combined-treatment initial bone density was normal (1·134 [SD 0·19] g/cm2 in group) had low granulocyte counts, the other two patients lumbar spine and 0·920 [0·14] g/cm2 in femoral neck).
Eight women in the combined-treatment group and one in the sulphasalazine group were using hormone replacement complaints, respectively. The remaining adverse events therapy. During the first 28 weeks, the mean changes in were not followed by withdrawal of study medication.
lumbar bone density change in the combined-treatment These included 18 patients (12 combined treatment) with and sulphasalazine group (n=64, 62) were Ϫ1·2% (Ϫ2·0, Ϫ0·3) and 0% (Ϫ0·9, 0·9) (p=0·06). At 56 weeks, the combined treatment) with gastrointestinal complaints (no changes were Ϫ1·3% (Ϫ2·3, Ϫ0·4) and Ϫ0·3% (Ϫ1·4, 0·8) ulcer or bleeding); ten (six combined treatment) with corresponding bone density changes were Ϫ0·6 (Ϫ2·1, 0·9) infarction; and eleven (five combined treatment) with skin versus Ϫ0·7 (Ϫ2·1, 0·7) over 28 weeks, and Ϫ1·9 (Ϫ3·1, Ϫ0·7) versus Ϫ1·3 (Ϫ2·5, Ϫ0·1) over 56 weeks (both laboratory abnormalities were reported in 37 cases (20 p>0·2). Eight versus six patients lost more than 5% (mean 8%) of spinal bone; 14 versus nine lost more than 5% Expected adverse effects such as weight gain and (mean 8%) of femoral neck bone. These losses typically hypertension were monitored at every control visit (ie, at occurred in the first 6 months, with stabilisation or least every 4 weeks); osteoporosis was assessed by twice- Lung function measurements showed no important changes during the first 56 weeks (data not shown).
Of patients who completed the 56-week treatment *One patient emigrated in week 40 of follow-up, one became pregnant.
†Treated with parenteral corticosteroids for pulmonary disease.
†Patients with baseline and at least one follow-up radiographic assessment.
‡Difference in total number of withdrawals: p=0·0008.
Table 3: Radiographic outcome of treatment Table 4: Reasons for withdrawal of patients from study withdrawal rate for any reason during the study period. The benefits of combined therapy on radiography persisted up to week 80. We believe, therefore, that this combined therapy can be classified as a disease-controllingantirheumatic therapy.
The differences between combined treatment and sulphasalazine are all the more striking becausesulphasalazine alone also performed well as a disease- controlling antirheumatic drug—the onset of action wasrapid, the withdrawal rate was low (observed 29%,expected 27–40%), and the radiological progression rate was similar to that in another study in early rheumatoidarthritis.2 The true clinical relevance of the effects of the combined treatment on radiological progression must beproven by follow-up studies. However, since demonstrationof any effect on radiological progression in rheumatoid arthritis is difficult, we believe that the decrease, to a third of that of sulphasalazine—and possibly a sixth of that of“symptom-modifying antirheumatic drugs”3 hydroxychloroquine2—will prove clinically relevant.
Figure 3: Effect of treatment on total radiographic damage Withdrawal rates for both toxicity and lack of efficacy were much lower with combined therapy than with Box-whisker plots of absolute radiographic damage scores (Sharp vander Heijde method; summed total scores for erosions and joint-space sulphasalazine, and expected side-effects (especially of narrowing in hands and feet). Horizontal line in box=median; limits of prednisolone such as osteoporosis) were of minor box=25th and 75th percentiles; whiskers=10th and 90th percentiles; importance. The low withdrawal rate may itself contribute values above and below these plotted separately.
to the differences in efficacy by preventing the loss ofantirheumatic effect during the time when treatment is sulphasalazine) were classified as non-compliant—eight switched from one agent to another. There was a slight were negative for sulphapyridine in at least one of the urine increase in infections with combined therapy but none led samples taken at weeks 16, 28, and 40; and one did not to (even temporary) protocol interruptions and all could be return medication for pill counts more than once. Fifteen treated without hospital admission. Such toxicity figures must be interpreted with caution; with its small sample size sulphasalazine) were classified as probably non-compliant.
and short follow-up, a clinical trial is not suited to reliable Thus, compliance was deemed satisfactory in 131 patients detection of side-effects that may be important. Given the (85%). Eight patients received intra-articular corticosteroid limited period and dose of the combination, we do not injections outside the permitted periods (two of them had expect important late morbidity. Nevertheless, the study been withdrawn; three were in combined-treatment group, cohort is being followed up so that long-term benefits and five in the sulphasalazine group). Eight patients in the risks of the therapy can be assessed. Thus we were able to read, analyse and report the 80-week radiography data corticosteroids after they were withdrawn from the trial. In ahead of time in response to criticism that 56 weeks of addition, one patient in the sulphasalazine group was withdrawn because she needed corticosteroids to treat The randomisation procedure of our trial created violations, contamination, and cointerventions were minor and did not affect the conclusions from intention-to-treat The combination of an extended oral pulse or analysis. Compliance was satisfactory, in an admittedly corticosteroids, methotrexate, and sulphasalazine led to an crude assessment. However, concealment of treatment immediate, substantial, and highly significant improvement allocation might raise some concerns. The treating of disease activity and physical function in most patients rheumatologists responsible for recruitment (but not with severe, early rheumatoid arthritis. On a low daily assessment) could potentially become aware of treatment maintenance dose of 7·5 mg prednisolone from week 7 onwards, this improvement continued at a slower pace up guaranteed that they had no clue to the treatment to week 28, and was almost double that of conventional allocation of subsequent patients. Outcome assessment was treatment with sulphasalazine alone in all clinical measures.
fully delegated to independent assessors: they were The study did not have sufficient power to show statistical unaware of the rapid effects of corticosteroids because they significance for differences in remission rates. Combined were not involved in the care of these patients, and assessed therapy also had beneficial effects on joint damage as them only at baseline and at weeks 16, 28, 40, and 56.
shown on hand and foot radiographs. This degree of They were asked not to discuss the disease or its treatment clinical efficacy could not be maintained by sulphasalazine with the patients. Since they were health professionals but alone—most of the clinical differences between the groups not physicians, they had less experience of corticosteroid decreased and were no longer significant after prednisolone side-effects. The patients were fully informed about the was withdrawn, and there were no further changes after potential side-effects of all the drugs. To protect against methotrexate was stopped. However, from inspection of an unblinding, they were told that response to treatment was area under the curve we suggest that the patients who variable in onset and efficacy with all three drugs. Although received combined treatment had major clinical benefit some unblinding due to large differences in efficacy cannot throughout the year. This view is supported by the low be ruled out, the effects of the subjective clinical measures were consistent, and reflected those in the objective withdrawn.26 By contrast, the effect of our combined treatment on the progression of joint destruction persisted Another concern that might be raised is the method of for up to 1 year after corticosteroids were stopped. The reading radiographs. To improve precision, these were read initially more rapid progression in the sulphasalazine group without knowledge of the patient’s identity or group slowed and approached that of the combined-treatment allocation, but they were ordered in sequence. This group with a lag of 1 year. These differences between our approach might be criticised as leading to bias. In general, study and that of Kirwan might be the result of more opinions are divided on the value of having previous effective treatment not only in our experimental group but information available when a judgment is made. In the case of radiographs in studies of rheumatoid arthritis, and especially in early disease, random-sequence reading temporary clinical effects of a prednisone regimen starting introduces a lot of noise, because small changes in with 15 mg daily. Other studies with step-down strategies positioning can temporarily hide erosions. In any case, both and oral or parental steroid pulses showing limited benefit Larsen’s original method20 and the Van der Heijde and troubling rebound effects.28–30 The limited clinical modification of Sharp’s method2 require sequential reading; efficacy of corticosteroids in other trials may be due to the these are the methods used in the majority of published use of doses inadequate to bring the disease under control studies (although some [eg, Kirwan5] did score randomly).
from the outset. Our results support the concept of step- In our trial, any bias would work in the same direction in down bridge therapy, and suggest that immediate and both groups, and would not alter the conclusions based on intensive suppression of high damage progression rates by a rapidly acting regimen may be sustained by another The eligibility criteria in our trial selected patients with regimen with a slower mode of onset (eg, sulphasalazine).
poor a-priori outlooks. For instance, a high proportion of 60 mg prednisolone is a high daily dose for rheumatoid patients had erosive disease at baseline (77%), partly arthritis. However, the rapid tapering schedule resulted in a because we included foot radiographs; on the basis of hand mean daily dose of 12 mg overall, and 7·5 mg daily from radiographs only the proportion of patients with erosive week 7 onwards. Also, the initial dose is not high compared disease would have been 47%. Nevertheless, some patients with standard (but not always proven) therapy for most included might not have needed such an intensive other severe autoimmune diseases (eg, myositis, vasculitis, approach as our combined treatment, since predictive lupus nephritis) and is much lower than the intravenous variables never provide 100% accuracy.21 On the other methylprednisolone pulses that have been tried with limited hand, such patients would probably show a good clinical arthritis. The tapering schedule after week 28 proved to be methotrexate would be possible; thus the risks involved in too rapid, resulting in partial loss of benefit and some their overtreatment are limited. We arbitrarily excluded disease flares. This factor may have contributed to the low patients older than 70, and further study is needed before numbers of lasting remissions. We look forward to further this therapy can be advised for patients over 70, or for studies with slower tapering schedules.
patients with a longer duration of rheumatoid arthritis or We chose methotrexate as the second drug in the concurrent disease. Thus, our results apply to otherwise combination because the onset of effect is rapid. However, healthy patients with early and active rheumatoid arthritis in this and another trial, the response to sulphasalazine was treated in a specialist setting. The opportunity to intervene as rapid as that to methotrexate.31 In the design phase, we early with second-line antirheumatic drugs relies heavily on set a fixed, low dose of methotrexate because little was the early diagnosis of rheumatoid arthritis and rapid known about the toxicity of our combination. We would, specifically organised to facilitate this process.
intensification of methotrexate or sulphasalazine in case of Combination therapy can be applied in many different suboptimum response (with the risk of additional toxic ways. We chose a step-down strategy with rapidly acting effects). The value of the combination of methotrexate and drugs. This approach optimised the chance of efficacy in a sulphasalazine is uncertain.31–33. In our study, tapering of potentially limited window of opportunity. Our results methotrexate had little impact on the mean results, support the view that corticosteroids are among the most although some patients had disease flares, which might be attributable to the withdrawal of prednisolone in the antirheumatic drugs. A meta-analysis has confirmed that preceding period. The contribution of 7·5 mg methotrexate low-dose corticosteroids can be beneficial.22 The toxic weekly to efficacy and toxicity was probably small in the effects of corticosteroids may be comparable to those of second 6 months of treatment. However, whether the some other disease-modifying antirheumatic drugs23 and efficacy of the full combination in the first 6 months can be even NSAIDs, and their use in established rheumatoid equalled by prednisolone/sulphasalazine or prednisolone/ arthritis is widespread. However, doubts about the methotrexate combinations remains to be seen. In view of longevity of the effects, and fears of cumulative morbidity the high benefit-risk ratio of our combined treatment in the have limited their use in early rheumatoid arthritis. The first 6 months, we believe maintenance of this ratio in the optimum dosing schedule is also unclear.
subsequent period is more desirable than reduction of the In 1959 a trial of prednisolone at daily doses of already low risk in the first 6 months.
10–20 mg indicated antirheumatic properties (both clinical To date, most drug combination trials have not shown and radiological), but with substantial side-effects.25 Kirwan addition or synergy (ie, benefits equal or better than the and colleagues5 showed an effect of 7·5 mg prednisone sum of benefits attributed to the single drugs). At best, daily on the development of radiological damage together investigators showed some enhanced efficacy at the expense with only temporary effect on disease activity measures, of some extra toxic effects. However, Tugwell and and no side-effects. These investigators have since reported that joint destruction recommences after corticosteroids are rheumatoid arthritis who had only a partial response to methotrexate showed clinically important improvement Association: a seven-day variability study of 499 patients with peripheral when cyclosporin was added to their regimen. O’Dell and rheumatoid arthritis. Arthritis Rheum 1965; 8: 302–34.
12 Jones E, Hanly JG, Mooney R, et al. Strength and function in the normal colleagues achieved 50% improvement in composite and rheumatoid hand. J Rheumatol 1991; 18: 1313–18.
symptoms of arthritis with methotrexate, sulphasalazine, 13 Tugwell P, Bombardier C, Buchanan WW, Goldsmith CH, Grace E, and hydroxychloroquine.34 80% and 50%, respectively, of Hanna B. The MACTAR patient preference disability questionnaire: an patients in these two studies were on low-dose individualized functional priority approach for assessing improvement in corticosteroid maintenance therapy. Both studies show that physical disability in clinical trials in rheumatoid arthritis. J Rheumatol
1987; 14: 446–51.
substantial improvement is possible in established disease, 14 Boers M, Tugwell P, Felson DT, et al. World Health Organisation and and that more study on the merits of combination therapy International League of Associations for Rheumatology core endpoints for symptom modifying antirheumatic drugs in rheumatoid arthritis Our study supports the view that corticosteroids in the clinical trials. J Rheumatol 1994; 41 (suppl): 86–89.
15 Siegert CEH, Vleming LJ, Vanderbroucke JP, Cats A. Measurement of proper regimen are among the most powerful disease- disability in Dutch rheumatoid arthritis patients. Clin Rheumatol 1984; 3:
controlling antirheumatic drugs available.
intensive intervention in rheumatoid arthritis with this 16 Pinals RS, Masi AT, Larsen RA. Preliminary criteria for clinical combination step-down schedule offers additional disease remission in rheumatoid arthritis. Arthritis Rheum 1981; 24: 1308–15.
control over an above that of sulphasalazine alone. Damage 17 Felson DT, Anderson JJ, Boers M, et al. American College of Rheumatology preliminary definition of improvement in rheumatoid control persists for up to 1 year after corticosteroids are arthritis. Arthritis Rheum 1995; 38: 727–35.
stopped, but to maintain optimum clinical efficacy after 28 18 Van der Heijde DMFM, van ’t Hof MA, Van riel PLCM, et al. Judging weeks, another tapering schedule is probably necessary.
disease activity in clinical practice in rheumatoid arthritis: first step in the
development of a disease activity score. Ann Rheum Dis 1990; 49:
All the authors contributed to design, analysis, and writing the report.
19 Van der Sluys Veer G, Soons JWPH. A time-resolved fluoroimmuno Maarten Boers was the principal investigator and study centre coordinator, assay of the IgM-rheumatoid factor. Eur J Clin Chem Clin Biochem 1992; Maastricht. Arco Verhoeven was trial coordinator, read radiographs, and 30: 301–05.
coordinated laboratory samples. Mart van de Laar was study centrecoordinator Enschede/Almelo, and coordinated rheumatoid factor and ANA 20 Larsen A, Dale K, Eck M. Radiographic evaluation of rheumatoid arthritis and related conditions by standard reference films. Acta Radiol Denderen, Derkjen van Zeben, Ben Dijkmans, André Peeters, Piet Jacobs, 1997; 18: 481–91.
and Hans van den Brink were study centre coordinators at their respective 21 van Zeben D, Hazes JMW, Zwinderman AH, Vandenbroucke JP, centres. Hubert Schouten coordinated statistical design and analysis and Breedveld FC. Factors predicting outcome of rheumatoid arthritis: was a member of the safety committee. Désirée van der Heijde did results of a follow-up study. J Rheumatol 1993; 20: 1288–96.
radiographic training and was quality supervisor. Annelies Boonen read 22 Saag KG, Criswell LA, Sems KM, Nettleman MD, Kolluri S. Low-dose radiographs. Sjef van der Linden was division chief.
corticosteroids in rheumatoid arthritis. Arthritis Rheum 1996; 39:
1818–25.
AcknowledgmentsThe COBRA trial was funded by Ontwikkelingsgeneeskunde, 23 Fries JF, Williams CA, Ramey D, Bloch DA. The relative toxicity of Ziekenfondsraad, The Netherlands (grant number 92-045). We thank disease-modifying antirheumatic drugs. Arthritis Rheum 1993; 36:
Pharmacia & Upjohn, Uppsala, Sweden, for providing sulphasalazine; the members of the safety committee A Beysens (clinical pharmacist) A Cats, 24 Morrison E, Capell H. Corticosteroids in the management of and H Rasker (rheumatologists); M van der Westerlaken and rheumatoid arthritis. Br J Rheumatol 1996; 35: 1–4.
F de Stoppelaar (clinical pharmacists) for preparation and distribution of the 25 Joint Committee of the Medical Research Council and Nuffield trial medication; M Braeken, L Heusschen-Houben, and all other research Foundation. A comparison of prednisolone with aspirin or other personnel; and all participating patients.
analgesics in the treatment of rheumatoid arthritis. Ann Rheum Dis 1959;
18: 173–87.
26 Kirwan JR, Hickling P, Jacoby R. Joint destruction recommences after glucocorticoids are withdrawn in early rheumatoid arthritis. Arthritis Harris ED Jr. Management of rheumatoid arthritis. In: Kelley WN, Rheum 1996; 39 (suppl): S218.
Harris ED, Ruddy S, Sledge CB, eds. Textbook of rheumatology.
27 van Schaardenburg D, Valkema R, Dijkmans BAC, et al. Prednisone Philadelphia: WB Saunders, 1993; 912–23.
treatment of elderly-onset rheumatoid arthritis; disease activity and bone van der Heijde DMFM, van Riel PL, Nuver-Zwart IH, Gribnau FW, van mass in comparison with chloroquine treatment. Arthritis Rheum 1995; de Putte LBA. Effects of hydroxychloroquine and sulphasalazine on 38: 334–42.
progression of joint damage in rheumatoid arthritis. Lancet 1989; i: 28 Ciconelli RM, Ferraz MB, Visioni RA, Oliviera, Atra E. A randomized double-blind controlled trial of sulphasalazine combined with pulses of Edmonds JP, Scott DL, Furst DE, Brooks P, Paulus HE. Antirheumatic methylprednisolone or placebo in the treatment of rheumatoid arthritis.
drugs: a proposed new classification. Arthritis Rheum 1993; 36: 336–39.
Br J Rheumatol 1996; 35: 150–54.
Van der Heide A, Jacobs JWG, Bijlsma JWJ, et al. The effectiveness of 29 van Gestel AM, Laan RFJM, Haagsma CJ, et al. Oral steroids as bridge early treatment with ‘second-line’ antirheumatic drugs, a randomised, therapy in RA patients starting with parental gold: a randomised double controlled trial. Ann Intern Med 1996; 124: 699–707.
blind placebo controlled trial. Br J Rheumatol 1995; 34: 347–51.
Kirwan JR. The effect of glucocorticoids on joint destruction in 30 Weusten BLAM, Jacobs JWG, Bijlsma JWJ. Corticosteroid pulse therapy rheumatoid arthritis. The Arthritis and Rheumatism Council Low-dose in active rheumatoid arthritis. Semin Arthritis Rheum 1993; 23: 183–92.
Glucocorticoid Study Group. N Engl J Med 1995; 333: 142–46.
31 Haagsma C, van de Putte L, van Riel P. Sulfasalazine, methotrexate and Tugwell P, Boers M. Long acting drug combinations in rheumatoidarthritis: updated overview. In: Wolfe F, Pincus T, eds. Rheumatoid combination in early rheumatoid arthritis, a double blind randomised arthritis: pathogenesis, assessment, outcome, and treatment. New York: study. Arthritis Rheum 1995; 38 (suppl): S368.
32 Haagsma CJ, van Riel PLCM, de Rooji DJRAM, et al. Combination of Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism methotrexate and sulphasalazine versus methotrexate alone. A Association revised criteria for the classification of rheumatoid arthritis.
randomised open clinical trial in rheumatoid arthritis patients resistant Arthritis Rheum 1988; 31: 315–24.
to sulphasalazine therapy. Br J Rheumatol 1994; 33: 1049–55.
Schouten HJA. Adaptive biased urn randomisation in small strata when 33 Dougados M, Cantagrel A, Goupille P, et al. Sulfasalazine, methotrexate, blinding is impossible. Biometrics 1995; 51: 1529–35.
and the combination in early rheumatoid arthritis: a double blind Smythe HA, Helewa A, Goldsmith CH. “Independent assessor” and randomized study. Arthritis Rheum 1996; 39 (suppl): S103.
“pooled index” as techniques for measuring treatment effects in 34 Tugwell P, Pincus T,Yocum D, et al. Combination with cyclosporine and rheumatoid arthritis. J Rheumatol 1977; 4: 144–52.
methotrexate in severe rheumatoid arthritis. N Engl J Med 1995; 333:
10 Anderson JJ, Felson DT, Meenan RF, Williams HJ. Which traditional measures should be used in rheumatoid arthritis clinical trials: Arthritis 35 O’Dell JR, Haire CE, Erikson N, et al. Treatment of rheumatoid arthritis Rheum 1989; 32: 1093–99.
with methotrexate alone, sulfasalazine and hydroxychloroquine, or a 11 The cooperating clinics committee of the American Rheumatism combination of all three medications. N Engl J Med 1996; 334: 1287–91.

Source: http://www.cobratherapie.nl/pdf/Boers%20et%20al%20Lancet%201997.pdf