Article Text

Download PDFPDF

Extended report
The comparative effectiveness of oral versus subcutaneous methotrexate for the treatment of early rheumatoid arthritis
  1. Glen S Hazlewood1,2,3,
  2. J Carter Thorne4,
  3. Janet E Pope5,
  4. Daming Lin6,
  5. Diane Tin4,
  6. Gilles Boire7,
  7. Boulos Haraoui8,
  8. Carol A Hitchon9,
  9. Edward C Keystone6,
  10. Shahin Jamal10,
  11. Vivian P Bykerk6,11
  12. for the CATCH Investigators
    1. 1Institute of Health, Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada
    2. 2Department of Medicine, University of Calgary, Calgary, Alberta, Canada
    3. 3McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, Alberta, Canada
    4. 4Southlake Regional Health Centre, Newmarket, Ontario, Canada
    5. 5Schulich School of Medicine and Dentistry, Western University, London, Ontario, Canada
    6. 6Mount Sinai Hospital, Toronto, Ontario, Canada
    7. 7Centre Hospitalier Universitaire de Sherbrooke (CHUS), Sherbrooke University, Sherbrooke, Quebec, Canada
    8. 8University of Montreal Hospital Research Centre (CRCHUM), Montreal, Quebec, Canada
    9. 10University of Manitoba, Winnipeg, Manitoba, Canada
    10. 10University of British Columbia, Vancouver, British Columbia, Canada
    11. 11Hospital for Special Surgery, New York, New York, USA
    1. Correspondence to Dr Vivian P Bykerk, Hospital for Special Surgery, 535 East 70th St., New York, NY 10021, USA; bykerkv{at}


    Objective To determine the comparative effectiveness of oral versus subcutaneous methotrexate (MTX) as initial therapy for patients with early rheumatoid arthritis (ERA).

    Methods Patients with ERA (symptoms ≤1 year) initiating MTX therapy were included from a multicentre, prospective cohort study. We compared the effectiveness between starting with oral versus subcutaneous MTX over the first year. Longitudinal multivariable models, adjusted for potential baseline and time-varying confounders, were used to compare treatment changes due to inefficacy or toxicity and treatment efficacy (Disease Activity Score-28 (DAS-28), DAS-28 remission and Health Assessment Questionnaire-Disability Index (HAQ-DI)).

    Results 666 patients were included (417 oral MTX, 249 subcutaneous MTX). Patients prescribed subcutaneous MTX were prescribed a higher dose of MTX (mean dose over first three months 22.3 mg vs 17.2 mg/week). At 1 year, 49% of patients initially treated with subcutaneous MTX had changed treatment compared with 77% treated with oral MTX. After adjusting for potential confounders, subcutaneous MTX was associated with a lower rate of treatment failure ((HR (95% CI) 0.55 (0.39 to 0.79)). Most treatment failures were due to inefficacy with no difference in failure due to toxicity. In multivariable models, subcutaneous MTX was also associated with lower average DAS-28 scores (mean difference (−0.38 (95% CI −0.64 to −0.10)) and a small difference in DAS-28 remission (OR 1.2 (95% CI 1.1 to 1.3)). There was no significant difference in sustained remission or HAQ-DI (p values 0.43 and 0.75).

    Conclusions Initial treatment with subcutaneous MTX was associated with lower rates of treatment changes, no difference in toxicity and some improvements in disease control versus oral MTX over the first year in patients with ERA.

    • Methotrexate
    • Early Rheumatoid Arthritis
    • Outcomes research
    View Full Text

    Statistics from


    Methotrexate (MTX) is widely regarded as the anchor drug of rheumatoid arthritis (RA) treatment with an established long-term safety profile and improved survival over time relative to other disease-modifying anti-rheumatic drugs (DMARDs).1–4 The use of MTX in the treatment of RA has evolved over time. Early trials and healthcare providers started with low doses (5–7.5 mg/week) and would escalate MTX using a slow-dose titration.5 Modern practice, however, includes the use of higher doses with rapid dose escalation.6 ,7 A systematic review of dosing practice of MTX found evidence of improved outcomes with higher doses.8

    The route of MTX delivery may be important. Absorption of MTX is improved with parenteral MTX, particularly at doses >15 mg/week.9 ,10 In a double-blind randomised controlled trial (RCT) that compared oral to parenteral administration, subcutaneous MTX (MTX) (15 mg/week) demonstrated superior efficacy to oral MTX (15 mg/week) but was associated with an increase in withdrawals due to adverse events.11 Observational studies of patients who switched from parenteral to oral MTX demonstrated high rates of flares in disease activity, with increased gastrointestinal side effects.12 ,13 A recent cohort study of patients taking subcutaneous MTX after oral MTX found high rates of treatment continuation, with 75% of patients still on treatment at 2 years.14 These observational studies, however, were limited by the lack of a comparator group.

    Comparative effectiveness research (CER) is a broad area of study aimed at determining the comparative benefits and harms of treatments in settings reflective of clinical practice and in assisting patients and physicians to choose between effective treatments.15 Observational research plays an important role in CER, as patient populations, treatment approaches and methods of evaluation may differ from clinical trials. The objective of this study was to compare the effectiveness of initial treatment with subcutaneous versus oral MTX for the treatment of early RA (ERA) in routine clinical care. Our primary research question was to determine whether patients starting subcutaneous MTX were more or less likely to require a change in treatment over the first year. We secondarily aimed to determine whether initial treatment with oral MTX or subcutaneous MTX was associated with improved effectiveness over the first year.


    The study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology checklist for cohort studies.16

    Study design and setting

    The Canadian Early Arthritis Cohort (CATCH) is an ongoing Canadian, multicentre, prospective, observational study of patients with ERA.17 To be eligible, patients must be >16 years old and have confirmed or suspected ERA. Patients are followed every three months in the first year and treated at the discretion of their rheumatologists. Comprehensive patient-reported and physician-reported measures and laboratory investigations are collected at each visit. For this study, patients were included from January 2007 to January 2014.


    The inclusion criteria for this study were a diagnosis of ERA according to 2010 American College of Rheumatology (ACR)/European League Against Rheumatism criteria,18 symptom duration of ≤1 year, treatment with MTX within the first three months of study entry with no or minimal prior exposure to MTX (minimal exposure defined as MTX for <4 weeks prior to study entry). Patients must have reached the first follow-up visit (3 months) and have at least one outcome (Disease Activity Score-28 (DAS-28)) available.


    The exposure was the first route used to administer MTX, either oral or subcutaneous. We compared treatment failure and treatment effectiveness between oral and subcutaneous MTX over the first year of follow-up. Treatment failure was defined as the need to escalate treatment with a new non-biological or biological DMARD, or a change in the dose or route of MTX due to inefficacy or side effects. An increase in the dose of MTX was only considered a failure after the first three months of therapy to allow for an initial upward titration of MTX dose. Patients who reduced therapy (discontinued a DMARD, lowered the dose of MTX or changed from subcutaneous MTX to oral MTX) without a side effect or an increased DAS-28 from the previous visit were not considered to have failed treatment. Efficacy outcomes were DAS-28-erythrocyte sedimentation rate (ESR) (four variable) (primary efficacy outcome),19 DAS-28 remission (DAS-28 <2.6), DAS-28 sustained remission (DAS-28 <2.6 on two consecutive visits) and Health Assessment Questionnaire-Disability Index (HAQ-DI).20

    Patient-level covariates that could potentially confound the association between route of MTX received and treatment response were measured (see table 1 for the list of all covariates). The dose of MTX at each visit was the average dose used within the 3-month visit window. Systemic corticosteroid use was separated into oral or intramuscular/intra-articular and recorded as yes or no at each visit. The use of other systemic and biological DMARDs was recorded at each visit.

    Table 1

    Demographic and clinical characteristics of patients at baseline

    Missing data

    Serological status was defined as positive (either rheumatoid factor (RF) or anti-citrullinated protein) or negative. If the baseline serological status (RF or anti-citrullinated protein antibody) was missing, we used the next closest observation, as conversion of baseline serological status is rare in patients with ERA.21 If the baseline erosion status was missing and there were no erosions on the next closest radiograph, then we assumed no erosions at baseline. If the baseline erosion status was missing and 3-month radiographs demonstrated erosions, we assumed erosions were present on the baseline radiographs. No imputation was performed for any other variable.

    Data analysis

    Baseline characteristics

    Baseline patient characteristics for patients initiating subcutaneous versus oral MTX were compared using χ2 tests for categorical variables, t tests for continuous variables and the Wilcoxon rank-sum test to compare the number of comorbidities.

    Comparative effectiveness

    Unadjusted treatment survival between initial subcutaneous and oral MTX was compared over the first year through Kaplan–Meier curves, with significance determined using the log-rank test. A Cox-proportional hazards model was used for the multivariable analysis of treatment failure. As our objective was to determine an association between our key exposure and the outcome, we used a full a priori model including all of the baseline covariates described. DAS-28, HAQ-DI and changes in DMARD therapy that were not part of the definition of a treatment failure (stopping a DMARD or lowering the dose of MTX without side effect) were modelled as time-varying covariates. All other covariates were modelled at baseline. The reasons for treatment failure were compared using χ2 tests.

    Multivariable analyses were performed for each clinical outcome (DAS-28, DAS-28 remission, HAQ-DI) using a mixed model to account for the repeated measurements of the outcome within individuals over time, while controlling for measured confounders.22 Time-varying covariates were changes in RA therapy (use of additional DMARDs, corticosteroid use, biological use, route of MTX after the initial visit, dose of MTX) and variables assessing disease activity that were not part of the outcome assessed (HAQ-DI for analyses of DAS-28 and DAS-28 for the analysis of HAQ-DI). The analysis was performed as a multivariable model with each covariate included as a separate variable. An unstructured correlation matrix was used, with optimal covariance structures assessed by comparing Quasi-Akaike Information Criterions across models.23

    Propensity score-adjusted models

    As a sensitivity analysis, we performed propensity score-adjusted models for treatment failure and DAS-28 (as a continuous variable). The propensity score included all baseline variables excluding the exposure (initial route of MTX) and time-dependent covariates. The propensity score was included as an untransformed covariate in the regression models.

    Effect of treatment centre variability

    The choice of route of MTX (subcutaneous or oral) was at the discretion of the treating rheumatologist. As such, treatment centre characteristics may potentially confound the association between medication choice and outcomes. To explore this, we performed several sensitivity analyses. First, we have previously shown an association between site size and treatment outcomes in the CATCH cohort.24 We therefore included a model with site size (number of patients from the treatment centre) included as a covariate. Second, we repeated the multivariable analyses for treatment failure and DAS-28, including only treatment centres in which at least 25% of patients used each route of MTX. Finally, for DAS-28, we included treatment centre as a random effect in the mixed model.

    All statistical analyses were performed in SAS, V.9.3 (SAS Institute, Cary, North Carolina, USA). All statistical tests were two-tailed and deemed to be statistically significant at α≤0.05.


    Patient population and baseline characteristics

    Of 2178 patients entered in the cohort, 666 met eligibility criteria and had at least one follow-up DAS-28 score over the first year (figure 1). Of the eligible patients, 586 (88%) started MTX right at or by the baseline visit, and 80 (12%) started by 3 months of entry into the study; 417 (63%) received initial therapy with oral MTX and 249 (37%) received subcutaneous MTX.

    Figure 1

    Flow chart of included and excluded patients. ACR. American College of Rheumatology; DAS-28, Disease Activity Score-28; EULAR, European League Against Rheumatism; MTX, methotrexate; RA, rheumatoid arthritis.

    Patients who were prescribed initial subcutaneous MTX were less likely to be co-prescribed other DMARDs, had marginally more comorbidities and were prescribed a higher dose of MTX (table 1). There was no significant overall difference in corticosteroid use, but patients prescribed subcutaneous MTX were less likely to receive oral corticosteroids and more likely to receive intramuscular or intra-articular corticosteroids. Patients prescribed subcutaneous MTX were also younger and had erosions reported more often on initial radiographs, although the result was not statistically significant. Other baseline characteristics, including disease activity, were similar. There were missing data for some baseline variables: serological status (9%), erosion status (14%), HAQ-DI (3%) and DAS-28 (7%). Other variables had no missing data.

    Comparative effectiveness

    In the unadjusted analysis, 49% of patients who were initially treated with subcutaneous MTX had changed treatment compared with 77% of patients with oral MTX (figure 2A). In the adjusted analysis, initial use of subcutaneous MTX was associated with a lower rate of treatment failure than oral MTX ((HR (95% CI) 0.55 (0.39 to 0.79)) (table 2). The use of intramuscular/intra-articular corticosteroids at any visit was also associated with a significantly increased rate of treatment failure (see online supplementary table S1), likely reflecting the common practice of co-administering intramuscular or intra-articular corticosteroids at the time of DMARD changes due to inefficacy. In total, 109 patients were censored prior to 1 year as they were withdrawn from the study. In a sensitivity analysis, when these patients were considered to have failed treatment, the association between subcutaneous MTX and treatment failure was similar (HR 0.61; 0.44 to 0.85).

    Table 2

    Adjusted association between route of initial methotrexate and outcomes

    Figure 2

    Unadjusted comparisons of initial subcutaneous versus oral methotrexate (MTX) over the first year for (A) survival of initial treatment strategy and (B) Disease Activity Score-28 scores.

    Most treatment failures were due to inefficacy, with a significant difference between subcutaneous and oral MTX (table 3). There was no difference in treatment failures between subcutaneous and oral MTX due to toxicity alone or due to combined inefficacy and toxicity (table 3).

    Table 3

    Treatment failure by reason, unadjusted (%)

    Treatment changes of patients at the time of failure are shown in table 4. Fewer patients prescribed initial subcutaneous MTX changed the route of administration, increased the dose of MTX after 3 months or added/switched a non-biological DMARD. More patients with subcutaneous MTX decreased the dose of MTX after 3 months, though the absolute percentages were low (10% with subcutaneous MTX compared with 4% with oral MTX). The percentage of patients starting a biologic over the first year was low, with no difference between groups.

    Table 4

    Treatment changes of patients at time of failure (%)

    DAS-28 scores decreased significantly over time to a greater extent with subcutaneous MTX than oral MTX at 3, 6 and 9 months, but not 12 months (figure 2B and online supplementary table S2). In the multivariable mixed model, subcutaneous MTX was independently associated with a greater reduction in average DAS-28 scores of 0.38 (95% CI 0.10 to 0.64) and a higher odds of DAS-28 remission (OR 1.2; 1.1 to 1.3) over the first year compared with oral MTX. The effect on DAS-28 remission was small, and there was no significant difference between subcutaneous and oral MTX for sustained remission (table 2). Patients who achieved DAS-28 remission with initial oral MTX at 1 year, however, required more treatment changes than patients treated with initial subcutaneous MTX (mean changes per patient 1.28 (SD 0.96) vs 0.84 (SD 0.79), p=0.003). There was no association between route of MTX exposure and HAQ-DI; both groups achieved significant improvements in HAQ-DI by 3 months with low mean values that persisted throughout follow-up (see online supplementary table S2).

    Propensity score-adjusted models

    In the propensity score-adjusted models, there remained a significant association between route of MTX and treatment failure (HR for treatment failure: 0.55; 0.39 to 0.79) and DAS-28 (mean difference (MD) −0.35; −0.61 to −0.09).

    Treatment centre variability

    There were 19 treatment centres, which included between 1 and 103 patients. In sites with >20 patients, the percentage of subcutaneous use varied from 0% to 96% (see online supplementary figure S1). After adjusting for centre size, subcutaneous MTX was still associated with a reduced risk of treatment failure (HR 0.52; 0.36 to 0.75) and lower DAS-28 scores over the first year (MD −0.40; −0.66 to −0.14). When we restricted the analysis to the four treatment centres in which 25% of patients were treated with both subcutaneous and oral MTX, there was no significant association between route of MTX and treatment failure (HR 0.65; 0.31 to 1.3) or DAS28 (MD −0.09; −0.49 to 0.31) among the 227 included patients. When treatment centre was included as a random effect in the mixed model, there was no association between subcutaneous MTX and DAS-28 (MD −0.14; −0.41 to 0.13).


    Our study demonstrated a significant association between initial subcutaneous administration of MTX and improved treatment continuation over the first year of treatment in patients with ERA compared with oral administration. The treatment failures were largely due to inefficacy, with no difference in treatment changes due to toxicity between oral and subcutaneous MTX. Subcutaneous MTX was also associated with lower mean DAS-28 scores over the first year and a small difference in achieving remission, but no difference in sustained remission or HAQ-DI. Many treatment centres expressed a strong preference for the route of delivery (either oral or subcutaneous). These treatment centres may differ in other unmeasured ways that affect outcomes and we were limited in our ability to control for this. While our findings suggest subcutaneous MTX is more effective than oral MTX as initial therapy, it is not possible to establish definitive causation given the observational nature of the study.

    Our findings were independent of the higher starting doses of MTX used in patients prescribed initial subcutaneous MTX. Most of the patients in our cohort received MTX at doses ≥15 mg/week. In this range, the bioavailability of oral MTX plateaus, whereas the bioavailability of subcutaneous MTX continues to increase.9 Thus, our findings may relate to the expected difference in bioavailable MTX with subcutaneous administration compared with oral administration at higher MTX doses.

    Strengths of our study include the multicentre design, prospective data collection and large sample size. We included rigorous methods of confounder control and examined outcomes using longitudinal models (which is preferred for cohort studies) as opposed to assessing outcomes only at a discrete time point. Prior observational studies have examined patient outcomes after switching from oral to parenteral MTX, or vice versa.12–14 These studies suggested improved outcomes and fewer side effects with parenteral administration, although all lacked a control group and were therefore not designed to assess the comparative effectiveness of oral versus subcutaneous MTX.

    The double-blind RCT of subcutaneous versus oral MTX by Braun et al11 also found improved outcomes (ACR20 response) with subcutaneous MTX, although the difference was small (78% vs 70% at 24 weeks). Our study complements these findings. While RCTs provide unbiased estimates of treatment efficacy by ensuring a random distribution of potential confounders, observational evidence provides data that may be more reflective of clinical practice. The starting dose of MTX in the RCT by Braun et al was 15 mg/week in both groups, lower than our study. The intervention in the RCT was also rigidly applied; patients who failed to achieve an ACR20 response by 16 weeks were required to change therapy. Moreover, the initial disease activity in our patient population was less severe compared with the study by Braun et al, with lower baseline DAS-28-4-ESR scores (5.5 vs 6.1–6.3) and HAQ-DI scores (1.1 vs 1.25–1.38).

    As with any cohort study, there is a potential for unmeasured confounding. Corticosteroid doses were not available for this analysis and could have been an important source of residual confounding. Differences between treatment centres may also be a source of unmeasured confounding. Bias may be introduced if physicians who prefer subcutaneous MTX differ in other ways from physicians who prefer oral MTX. We measured and adjusted for RA medications including corticosteroids, DMARDs and biologics at baseline and over time. Thus, the effect seen in our study was independent of any differences between treatment centres in the use of DMARDs at baseline or over time. It is possible that other factors such as medication adherence or response to non-pharmacological interventions are improved in sites that use higher rates of subcutaneous MTX. Physicians who prefer subcutaneous MTX may also be more likely to persist with prescribing this route of therapy, although likewise this bias may exist for physicians with a strong preference for oral MTX. The ability to control for an additional effect from treatment centre differences was limited due to the low within-centre variability in the choice of MTX route. The large magnitude of effect (particularly for treatment continuation) after adjusting for all measured covariates, however, increases the strength of our findings.25

    There may be other reasons that oral MTX is prescribed preferentially over parenteral administration. Patient preference should guide therapy, although the large variability observed among treatment centres suggests physicians largely determine treatment choice. There is also little evidence to suggest that patients prefer oral over parenteral MTX. Parenteral formulations have been shown to have high tolerability.26 Cost may also be an issue, although generic subcutaneous MTX, available in many countries, is roughly 10% the cost of biological therapy;27 optimising the use of traditional DMARDs may have cost savings.

    Whether oral or subcutaneous MTX should be the first treatment choice warrants further study, there is certainly little evidence to support oral MTX as the preferred route of delivery, as has been previously recommended.8 A useful next step would be a pragmatic RCT, where patients are randomised to oral versus subcutaneous MTX, with flexible application of the intervention and liberal inclusion criteria reflective of clinical practice. Until additional data are available, our findings, when added to the existing literature, support the use of subcutaneous MTX as the preferred mode of MTX administration.


    View Abstract

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


    • Handling editor Tore K Kvien

    • Collaborators The CATCH investigators include the following: Memorial University: Dr Majed Khraishi; Université de Sherbrooke: Dr Gilles Boire, Director of Ethical Scientific Conduct; Institut de Rhumatologie: Dr Boulos Haraoui, Vice Director, Dr Edith Villeneuve; McGill University: Dr Murray Baron, Dr Ines Colmegna, Dr Susan Bartlett; Hôpital Maisonneuve Rosemount: Dr Michel Zummer; University of Toronto: Dr Vivian Bykerk, Director and Principal Investigator, Dr Ed Keystone, Senior Consultant, Dr Pooneh Akhavan, Dr Lawrence Rubin, Dr Bindee Kuriya; Southlake Regional Health Centre: Dr Carter Thorne, Director of Operations, Ms. Diane Tin—AHP; Headwaters Health Care Centre: Dr Vandana Ahluwahlia; McMaster University, Hamilton, Ontario: Dr William Bensen, Dr Maggie Larche; St. Joseph's Health Centre, London: Dr Janet Pope, Research Director, Dr Lillian Barra; University of Manitoba: Dr Carol Hitchon, Director of Ethical Scientific Conduct; University of Saskatchewan: Dr Bindu Nair; Calgary Health Sciences Centre: Dr Christopher Penney, Dr Dianne Mosher, Dr Cheryl Barnabe, Dr Glen Hazlewood; Lethbridge, Alberta: Dr Hector Arbillaga; Grande Prairie, Alberta: Dr Christopher Lyddell; Vancouver Coastal Health: Dr Alice Klinkhoff, Dr Shahin Jamal; management Team: Ms. Franci Sniderman; statistical Support: Daming Lin, Jim Wang, McDougall Scientific.

    • Contributors All authors except DL were involved with patient recruitment. GH, JCT, JP, DL and VB developed the study plan. GH, DL, VB completed the data analysis. GH and VB wrote the draft manuscript. All authors reviewed, revised and approved the final manuscript.

    • Funding The CATCH study was designed and implemented by the investigators. The authors have received unrestricted grants from: Amgen and Pfizer Canada—Founding sponsor since 2007, Hoffmann-LaRoche, UCB Canada, Bristol-Myers Squibb Canada, AbbVie Corporation and Janssen Biotech since 2011. Dr Bykerk receives funding from the NIH.

    • Competing interests VB was asked to be a consultant for Medac and Antares (after this study was conceived, designed and analysed). GH is supported by an Alberta Innovates Health Solutions Clinical Fellowship. VB is supported by the Cedar Hill Foundation, New York, NY.

    • Ethics approval Local ethics committees.

    • Patient consent Obtained.

    • 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.