Article Text

Extended report
Predictors of response to methotrexate in early DMARD naïve rheumatoid arthritis: results from the initial open-label phase of the SWEFOT trial
  1. Saedis Saevarsdottir1,2,
  2. Helena Wallin1,
  3. Maria Seddighzadeh1,
  4. Sofia Ernestam1,
  5. Pierre Geborek3,
  6. Ingemar F Petersson4,
  7. Johan Bratt1,
  8. Ronald F van Vollenhoven1,
  9. for the SWEFOT Trial Investigators Group
  1. 1Rheumatology Unit, Department of Medicine, The Karolinska Institute and Karolinska University Hospital, Stockholm, Sweden
  2. 2Institute of Environmental Medicine, The Karolinska Institute, Stockholm, Sweden
  3. 3Department of Rheumatology, Lund University Hospital, Lund, Sweden
  4. 4Department of Orthopedics, South Sweden Musculoskeletal Research Centre, Lund University Hospital, Lund, Sweden
  1. Correspondence to Dr Saedis Saevarsdottir, Rheumatology Unit, Department of Medicine, Karolinska University Hospital, 17176 Stockholm, Sweden; saedis.saevarsdottir{at}


Objective To identify predictors of response to methotrexate (MTX) in early rheumatoid arthritis (RA).

Methods In the SWEFOT trial, patients with RA with symptom duration <1 year started MTX monotherapy (20 mg/weekly) and 405/487 continued until the 3–4- month visit. The primary outcome measure was the DAS28-based European League against Rheumatism (EULAR) response criteria. Multivariate logistic regression was used to study the association between response and the following baseline characteristics: gender, age, symptom duration, cigarette smoking habits, autoantibody status, Health Assessment Questionnaire (HAQ) score, concurrent prednisolone and treatment with non-steroidal anti-inflammatory drugs. Secondary response and remission measures were the American College of Rheumatology and the Simple Disease Activity Index and Clinical Disease Activity Index (SDAI/CDAI)-derived criteria.

Results After 3–4 months of MTX treatment, the frequency of EULAR good/moderate/no response was 34%/41%/25%, respectively. Parameters associated with a decreased likelihood of EULAR response were female gender (adjusted OR (adj OR) 0.50, 95% CI 0.31 to 0.81), symptom duration (adj OR per month increase 0.93, 95% CI 0.88 to 0.99), current smoking (adj OR 0.35, 95% CI 0.20 to 0.63) and higher HAQ (adj OR 0.56, 95% CI 0.40 to 0.80). Parameters associated with an increased likelihood of EULAR response were higher age (adj OR per 10-year increase 1.30, 95% CI 1.11 to 1.51) and prednisolone treatment (adj OR 2.84, 95% CI 1.43 to 5.63). The findings were similar when patients on prednisolone were excluded and other response criteria tested, although current smoking was the only significant predictor using all response criteria, while HAQ was the only significant predictor of all the remission criteria used. A matrix showed up to ninefold differences between subgroups stratified by the main predictors.

Conclusion Current smoking, female sex, longer symptom duration and younger age predict a worse response to MTX in patients with new-onset RA.

TrialRegNo NCT00764725

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.


Methotrexate (MTX) has been the first-line disease-modifying antirheumatic drug (DMARD) for rheumatoid arthritis (RA) since the 1980s, having the best documented efficacy and fewest adverse effects.1 2 However, only about one-third of patients respond well to MTX and some need to terminate their treatment because of adverse effects.3,,5 For patients who do not respond to MTX, the addition of tumour necrosis factor (TNF) inhibitors was shown to be superior to triple therapy with sulfasalazine and hydroxychloroquine in the SWEdish PharmacOTherapy (SWEFOT) trial.6 Furthermore, initial combination treatment with MTX and a TNF inhibitor3 4 7 8 or MTX and glucocorticoids8 has been reported to be more effective than MTX in monotherapy. Several other potent expensive immunologically-designed biological agents have emerged during the last decade.9

In order to choose the initial treatment in a rational manner, baseline parameters are needed that can help predict which patients will not respond to MTX. These patients could receive more potent treatment early, avoiding delays in finding the optimum treatment for inflammatory control that could lead to irreversible damage.10,,12 As disease activity at the commonly used 3–4-month follow-up visit strongly correlates with that after 1 year, this is a useful timepoint for deciding whether treatment changes are needed to achieve early inflammatory control.13 14 Furthermore, disease duration, disability and failing with a previous DMARD have been associated in a meta-analysis with less likelihood of response.15

Several studies have tried to outline predictive factors in RA16, some of which have been limited to predictors of MTX response. In early RA (BeSt trial, n=205), female sex, high baseline disease activity and high Health Assessment Questionnaire disability (HAQ) score were associated with less likelihood of achieving low disease activity following MTX monotherapy.17 Female sex and high baseline disease activity were also associated with less efficacy of MTX in a study on established RA, as was lack of non-steroidal anti-inflammatory drug (NSAID) treatment.18

Modifiable predictors for MTX response would, of course, be of particular interest. Recently, data from the EIRA study showed that current smoking was associated with less response to MTX in early RA (n=873).19 A similar non-significant trend was observed in the BeSt trial.17

The design and size of the standard care-based SWEFOT trial,6 where all patients initially received open-label MTX monotherapy, provided a further opportunity for simultaneous evaluation of baseline clinical and serological parameters (rheumatoid factor (RF), anti-cyclic citrullinated peptide antibody (anti-CCP)) and also smoking habits as potential predictors of MTX response in patients with DMARD-naïve early RA.


Study population: the SWEFOT trial

The SWEFOT trial was designed to compare two treatment strategies in patients with new-onset RA who had an incomplete response after an initial 3–4-month treatment period with MTX monotherapy and folic acid supplementation: (1) DMARD combination by adding sulfasalazine (SSZ) and hydroxychloroquine (HCQ); and (2) addition of a TNF inhibitor. Thus, all patients started treatment with MTX at an initial dose of 10 mg weekly which was increased biweekly by 5 mg increments to 20 mg weekly. This strategy allowed identification of patients who were not in need of more intensive therapy on clinical grounds, while those with an incomplete response were included in the randomisation. The study has been described in detail elsewhere.6

Fifteen rheumatology units in Sweden collaborated in the SWEFOT trial. Inclusion criteria were age >18 years, fulfilling classification criteria for RA according to the American Rheumatism Association 1987 revised criteria20 and symptom duration of <1 year (n=487). Exclusion criteria were contraindications to any of the trial medications or previous DMARD treatment. Patients who received glucocorticoid therapy had to have been on a stable dose for at least 4 weeks of at most 10 mg daily prednisolone (or equivalent) before study entry and throughout the study period. The disease activity was measured by a rheumatologist using the 28-joint count score (DAS28)21 and a score of >3.2 was required to enter the trial.

Patients eligible for analyses

A total of 487 patients entered the study. Of these, 27 stopped MTX treatment before the 3–4-month follow-up due to side effects, 46 for other reasons and 9 patients were diagnosed with another disease in the interim. Thus, 405 patients were eligible for the as-observed response analyses. In a sensitivity analysis the patients who discontinued MTX prior to the 3–4-month follow-up were included as non-responders.

Baseline variables

The following baseline parameters were evaluated, in addition to the disease characteristics: age, gender, symptom duration and smoking status (never, past or current cigarette smokers). Concurrent low-dose prednisolone treatment and regular NSAID use were also registered. Analysis of anti-CCP antibodies was made with the standard ELISA (Immunoscan-RA Mark2 ELISA test, Euro-Diagnostica, Malmö, Sweden).

Measures of disease activity

The disease activity was measured by DAS28 based on erythrocyte sedimentation rate (ESR), but was replaced by C reactive protein for five patients where ESR was missing.21 As secondary measures, the Simple Disease Activity Index (SDAI) and the Clinical Disease Activity Index (CDAI) were calculated (see table 1 in online supplement for formulae and categories of disease activity).22 23

Definition of response to treatment

Evaluation of the response to MTX was done at the 3–4- month follow-up visit. Those patients who had discontinued their MTX therapy prior to this visit were, regardless of reason, excluded from the principal as-observed analyses but were included as non-responders in a subsequent sensitivity analysis. In accordance with recent joint guidelines from the European League against Rheumatism (EULAR) and the American College of Rheumatology (ACR),24 the EULAR response criteria were used as the primary outcome measure. As secondary outcome measures, both the ACR criteria for ACR20, ACR50 and ACR70 response and also the SDAI and CDAI response criteria were used (see table 1 in online supplement).22 25,,29

Statistical analysis

Correlations were assessed by the Spearman correlation test. Predictor analyses were undertaken using logistic regression and the results were expressed as ORs with 95% CI. The primary outcome measure was EULAR overall response and the association between baseline variables and the likelihood of achieving a higher EULAR response category (no/moderate/good response) was evaluated using univariate and multivariate ordinal logistic regression. Baseline variables considered in the model, chosen based on correlation and clinical relevance, were age (in 10-year increments), gender, symptom duration (per month increase), anti-CCP antibody status (positive/negative), RF status (positive/negative), DAS28 and HAQ scores, concurrent corticosteroid use (yes/no) and concurrent daily NSAID use (yes/no). Anti-CCP and RF status showed high collinearity (r=0.31) so only anti-CCP status was kept in the model and excluded when RF status was analysed. Similarly, HAQ score and DAS28 showed high collinearity (r=0.51) and therefore only HAQ score was kept in the model and excluded when DAS28 was analysed.

To evaluate the consistency of the findings, the association between the identified predictors and response was analysed in a similar manner with multivariate logistic regression using other response measures—that is, the ACR20, ACR50, ACR70, SDAI- and CDAI-based (no/minor/major response) response criteria, as well as for remission as defined on the basis of DAS28, SDAI and CDAI (see table 1 in online supplement).

Statistical analysis was performed using the SAS 9.1 software (SAS Institute Inc, Cary, North Carolina, USA). All tests were two-sided and the significance level was set at 0.05.


In table 1 the baseline characteristics are shown for the 405 patients that were eligible for assessment. These patients represent a fairly typical population of newly diagnosed patients with RA. Of these patients, 395 had response data and a full set of potential baseline predictors was available for 322, but information about smoking habits, RF and anti-CCP antibody status was missing for 51, 50 and 18 patients, respectively. This group did not differ from the overall group in the baseline characteristics.

Table 1

Baseline characteristics of patients in SWEFOT study treated with methotrexate in monotherapy

Table 2 shows the proportions of patients who, at the 3–4- month follow-up visit, achieved the response criteria and remission according to EULAR, ACR, SDAI and CDAI. The EULAR response criteria were chosen as the primary outcome measure and 34% were EULAR good responders, 41% were moderate responders and 25% were non-responders. The proportion of responders according to ACR50 and SDAI major response was similar to that of EULAR good response, while the proportion achieving CDAI major response was somewhat higher. Non-responders according to EULAR, SDAI and CDAI were of similar frequency. Fewer patients achieved remission according to SDAI and CDAI criteria than EULAR criteria.

Table 2

Proportion of patients reaching each category of different response and remission criteria after 3–4 months of methotrexate monotherapy

Predictors of EULAR response

The influence of each baseline parameter on the likelihood of reaching a higher EULAR response category was evaluated by ordinal logistic regression. As shown in table 3, the findings from the univariate and multivariate analyses did not differ significantly for each potential predictor, neither did the univariate associations in this complete set differ from those found in the overall group with available data on each potential predictor (data not shown).

Table 3

Predictors of higher EULAR response at 3-month follow-up to MTX monotherapy*

Female sex was strongly associated with a decreased likelihood of responding to MTX (adjusted OR 0.50, 95% CI 0.31 to 0.81). Duration of symptoms before start of treatment was marginally associated with less response (adjusted OR per month increase 0.93, 95% CI 0.88 to 0.99). Current cigarette smoking was strongly associated with less likelihood of responding (adjusted OR 0.35, 95% CI 0.20 to 0.63), while prior smoking was not significantly associated with response (adjusted OR 0.68, 95% CI 0.41 to 1.12). A higher HAQ score was also associated with less likelihood of response (adjusted OR 0.56, 95% CI 0.40 to 0.80). Higher DAS28 at start of treatment was also associated with a decreased likelihood of response (adjusted OR 0.64, 95% CI 0.52 to 0.80), but it should be kept in mind that the EULAR response criteria are based on the change in DAS28. In contrast, concurrent treatment with stable low-dose prednisolone was associated with an increased likelihood of response (adjusted OR 2.84, 95% CI 1.43 to 5.63), as was higher age (adjusted OR per 10-year increase 1.30, 95% CI 1.11 to 1.51). The presence of RF or anti-CCP antibodies did not influence the response.

In our analyses we excluded 73 patients who stopped MTX treatment before the 3-month follow-up due to side effects or other reasons, and 33 of these had a complete dataset for multivariate analyses. In an intention-to-treat sensitivity analysis, these patients were included as non-responders; this resulted in very similar findings (data not shown).

Patients without glucocorticoid treatment

As concurrent treatment with low-dose prednisolone was strongly associated with response, the multivariate analysis was also performed for those without prednisolone treatment (N=284). All findings remained in this subgroup (figure 1); female gender, current smoking, longer duration and more disability predicted a lower EULAR response while higher age predicted a higher EULAR response.

Figure 1

Gender, baseline age, current smoking and disability (HAQ score) as predictors for reaching different outcome measures at the 3–4-month follow-up visit in patients without prednisolone treatment: (A) higher EULAR response category; (B) higher SDAI response category; (C) higher CDAI response category; (D) ACR20 response. The results are expressed as adjusted OR with 95% CI calculated using multivariate logistic regression. The construction of the models and definition of co-variables is further described in the Methods section. ACR, American College of Rheumatology; CDAI, Clinical Disease Activity Index; EULAR, European League against Rheumatism; HAQ, Health Assessment Questionnaire; SDAI, Simple Disease Activity Index.

The findings did not differ significantly between women and men, between current and non-smokers or between anti-CCP-positive and anti-CCP-negative patients (data not shown).

Validation of the findings with other response measures

To study the consistency of the primary outcome findings, the SDAI, CDAI and ACR criteria were applied. The findings are summarised in figure 1, excluding the few (14%) patients receiving prednisolone treatment, as the main aim of the study was to identify predictors of MTX monotherapy, not combination therapy. The results in the whole group did not differ from those presented (data not shown).

At the 3–4-month follow-up visit, 54% had reached ACR20 response, 33% had reached ACR50 response and 13% fulfilled the criteria for ACR70 response. The corresponding figures for SDAI response were 33% major, 35% minor and 32% no response and for CDAI responders were 53% major, 22% minor and 25% no response (table 2).

Female gender tended to be associated with less likelihood of response as measured by the SDAI (figure 1B), CDAI (figure 1C), ACR20 (figure 1D), ACR50 (adjusted OR 0.64, 95% CI 0.36 to 1.15) and ACR70 (adjusted OR 0.47, 95% CI 0.20 to 1.08) response criteria.

Higher age was significantly associated with an increased likelihood of response as defined by CDAI (figure 1C), and similar but non-significant trends were observed when the criteria for SDAI (figure 1B), ACR20 (figure 1D), ACR50 (adjusted OR 1.16, 95% CI 0.96 to 1.41), and ACR70 (adjusted OR 1.20, 95% CI 0.89 to 1.63) response were applied.

Current smoking was significantly associated with less likelihood of response as defined by the SDAI, CDAI and ACR20 response criteria (figure 1B and C), but non-significant findings were observed when the ACR50 (adjusted OR 1.16, 95% CI 0.96 to 1.41) and ACR70 (adjusted OR 0.74, 95% CI 0.24 to 2.26) response criteria were applied.

The findings for HAQ and the other analysed variables were non-significant when the response criteria of SDAI, CDAI and ACR were applied.

Predictors of remission

Being the ultimate goal of treatment, remission was also evaluated as a secondary outcome measure, applying the definitions derived from the DAS28, SDAI and CDAI disease activity measures (see table 1 in online supplement). This is presented for the patients not receiving prednisolone (figure 2), but similar findings were observed in the whole group (data not shown). Owing to a smaller proportion achieving remission, the statistical power to detect differences was less than for the response criteria above. Only higher HAQ was significantly associated with less likelihood of remission, while the findings for female gender, smoking and age were non-significant, although in the same direction as for the response measures. In addition, the baseline value of the disease activity measure from which each remission definition was derived was significantly associated with the likelihood of reaching that remission definition after 3–4 months.

Figure 2

Gender, baseline age, current smoking and disability (HAQ score) as predictors for reaching remission at the 3–4-month follow-up visit in patients without prednisolone treatment, as defined by (A) EULAR (DAS28 <2.6); (B) SDAI (≤2.8) and (C) CDAI ≤3.3). The results are expressed as adjusted OR with 95% CI, calculated using multivariate logistic regression. The construction of the models and definition of co-variables is further described in the Methods section. CDAI, Clinical Disease Activity Index; DAS28, 28-joint Disease Activity Score; EULAR, European League against Rheumatism; HAQ, Health Assessment Questionnaire; SDAI, Simple Disease Activity Index.

Prediction matrix

A matrix based on the main predictors (gender, smoking status and HAQ in tertiles) was constructed to visualise the actual findings, showing the proportion with EULAR non-response (figure 3, top), good response (middle) and remission (bottom). Up to ninefold differences were observed. Further subgrouping by age tertiles highlighted these differences even more, but the numbers in some subgroups became small (data not shown). For example, 71% of non-smoking men aged >60 years (highest age tertile) were good responders and there were no non-responders, while no women aged <50 years (lowest tertile) who smoked achieved a good response and 50% were non-responders.

Figure 3

Matrix showing the proportion (in percentages) with EULAR non-response (top), good response (middle) and remission (bottom) to methotrexate (MTX) monotherapy after 3–4 months in subgroups of patients with early rheumatoid arthritis (RA) in the SWEFOT trial. DAS28, 28-joint Disease Activity Score; EULAR, European League against Rheumatism.


In this standard care-based trial on patients with early RA, current smoking, female gender, lower age and more disability or disease activity at treatment start predicted less likelihood of a response to MTX treatment. This was observed using the EULAR criteria as the response measure, and validation with the ACR20, SDAI and CDAI criteria yielded similar findings overall, although current smoking was the only predictor showing a significant association using all four response measures. The power for identifying significant predictors of remission was smaller, and only the disability score (HAQ) and the baseline disease activity measures were significant predictors using all remission definitions. However, a risk matrix subgrouped by gender, smoking status and HAQ showed a gradient with manifold differences between patients having all or none of those risk factors (proportion with EULAR non-response 52% vs 6%, good response 12% vs 69%, remission 8% vs 50%, respectively).

For current smoking as a negative predictor of therapeutic efficacy, our findings are in accordance with our recent findings in the EIRA study.19 One smaller study in early RA reported a non-significant association between smoking and low disease activity following MTX treatment,17 and other studies on early RA not restricted to MTX have found more DMARD use in smokers,30 while findings on the influence of smoking on disease activity have been somewhat inconsistent.31 32

Our finding that female gender and high disability and disease activity at baseline are associated with less response to MTX treatment are also in accordance with previous studies in early17 and established RA.18 In addition, we found that higher age was associated with MTX response.17

Patients on stable low-dose prednisolone before and during the follow-up were allowed to participate. Still, the small group of patients on concurrent prednisolone treatment (14%) responded better. This is in accordance with trials where the combination of glucocorticoids and DMARD in early RA is superior to DMARD monotherapy, but in these studies the glucocorticoids were added at the same time.8 33,,35 As the response to prednisolone is more rapid than with MTX, it is expected that it will already have occurred at baseline. However, to eliminate the potential influence of this effective treatment, multivariate analyses were also performed in non-prednisolone-treated patients only and all findings remained in this subgroup.

Neither anti-CCP status nor RF status influenced the response. This is in line with one early RA study17 and a recent report on inflammatory polyarthritis,36 and illustrates that the predictors of clinical response and radiographic damage differ as anti-CCP and RF positivity has been associated with radiographic progression.37

The strength of this study is that it is conducted within a clinical trial which was performed as part of standard care in an unselected population of patients with RA. Furthermore, this is the largest study to date, which enabled simultaneous evaluation of these factors and also subanalyses and matrix construction of clinically important subsets. Another strength is that we tested the consistency of the findings with different response and remission criteria sets (derived from EULAR, ACR, SDAI, CDAI). In the trial setting, response criteria are regarded as the outcome measure of choice as they denote the change between two time points while remission criteria have low sensitivity to change.28 This was indeed observed in the present study. However, as remission criteria are easier to apply in routine care and are the ultimate goal of treatment, it was of interest also to evaluate the main predictors identified against the remission outcome. Only HAQ and the disease activity measured used showed significant associations, but striking differences were observed in matrix subgroup analyses for remission, just as for the response groups (figure 3).

Owing to the trial design, it was not possible to evaluate predictors of MTX response after 3–4 months of follow-up. However, this is a usual time frame before evaluating response in clinical practice, as early inflammatory control is important and many efficient treatment options are available. Furthermore, disease activity at this time point is strongly associated with disease activity after 6 or 12 months.13 14 Another limitation of this study is that information about smoking habits and serology was not available for all patients, which might give rise to biased results in the multivariate analyses. However, baseline characteristics were not different in these patients compared with the whole group, and the univariate analyses in this group did not differ from the whole group. The number of patients with adverse effects was too low for meaningful analysis of predictors.

In conclusion, our findings in this large standard care-based study of early RA provide new evidence that non-smoking is a modifiable risk factor which significantly improves the chance of responding to MTX. We also confirmed that female sex, lower age and more disability are major non-modifiable predictive factors for a poor response to MTX in early RA. In addition, patients on stable low-dose prednisolone when starting MTX responded better. Both modifiable and non-modifiable factors must be taken into account when choosing the optimal therapeutic strategy for our patients. Thus, these standard care-derived prediction aspects should be helpful both in daily clinical practice where they can be checked further when other aspects might play a role, and in further studies aimed at identifying additional response predictors and algorithms.


The authors thank all participating patients as well as the study nurses, co-investigators and colleagues who made this trial possible. The authors also thank Professor Lars Klareskog for critical reading of the manuscript and constructive comments.


Supplementary materials

  • Web only data

    Files in this Data Supplement:


  • NCT00764725 (

  • SWEFOT Trial Investigators Group Lars Cöster, Linköping; Eva Waltbrand, Borås; Agneta Zickert, Stockholm; Jan Theander, Kristianstad; Åke Thörner, Eskilstuna; Helena Hellström, Falun; Annika Teleman, Halmstad; Christina Dackhammar, Mölndal; Finn Akre, Örebro; Kristina Forslind, Helsingborg; Lotta Ljung, Umeå; Rolf Oding, Västerås; Katerina Chatzidionysiou, Stockholm; Margareta Wörnert, Stockholm.

  • Funding The study was supported in part by a grant from the Swedish Rheumatism Association. SS, RvV and SE were supported by clinical research funds from Stockholm county (ALF funds). An annual unrestricted grant was provided by Schering-Plough Sweden that was used to support the study coordinator (MW) and a medical monitor.

  • Competing interests None.

  • Patient consent All patients were provided with oral and written information prior to inclusion and consented to participate by signing the informed consent document.

  • Ethics approval This study was conducted with the approval of the regional ethical committees of the participating clinics in Sweden.

  • Provenance and peer review Commissioned; externally peer-reviewed.