Methotrexate for juvenile idiopathic arthritis

Abstract

Background

Juvenile idiopathic arthritis (JIA) is the most common rheumatic disease in childhood. Methotrexate has broad immunomodulatory properties and is the most commonly used disease‐modifying antirheumatic drug (DMARD). This is an update of a 2001 Cochrane review. It supports a living guideline for children and young people with JIA.

Objectives

To assess the benefits and harms of methotrexate for children and young people with juvenile idiopathic arthritis.

Search methods

The Australian JIA Living Guideline Working Group created a registry of all randomised controlled trials (RCTs) of JIA by searching CENTRAL, MEDLINE, Embase, and trials registries. The date of the most recent search of online databases was 1 February 2023.

Selection criteria

We searched for RCTs that compared methotrexate with placebo, no treatment, or another DMARD (with or without concomitant therapies) in children and young people (aged up to 18 years) with JIA.

Data collection and analysis

We used standard Cochrane methods. The main comparison was methotrexate versus placebo. Our outcomes were treatment response, sustained clinically inactive disease, function, pain, participant global assessment of well‐being, serious adverse events, and withdrawals due to adverse events. We used GRADE to assess the certainty of evidence for each outcome.

Main results

We identified three new trials in this update, bringing the total number of included RCTs to five (575 participants). Three trials evaluated oral methotrexate versus placebo, one evaluated methotrexate plus intra‐articular glucocorticoid (IAGC) therapy versus IAGC therapy alone, and one evaluated methotrexate versus leflunomide. Doses of methotrexate ranged from 5 mg/m²/week to 15 mg/m²/week in four trials, and participants in the methotrexate group of the remaining trial received 0.5 mg/kg/week. Trial size varied from 31 to 226 participants. The average age of participants ranged from four to 10 years. Most participants were females and most had nonsystemic JIA. The study that evaluated methotrexate plus IAGC therapy versus IAGC therapy alone recruited children and young people with the oligoarticular disease subtype of JIA.

Two placebo‐controlled trials and the trial of methotrexate versus leflunomide were adequately randomised and blinded, and likely not susceptible to important biases. One placebo‐controlled trial may have been susceptible to selection bias due to lack of adequate reporting of randomisation methods. The trial investigating the addition of methotrexate to IAGC therapy was susceptible to performance and detection biases.

Methotrexate versus placebo

Methotrexate compared with placebo may increase the number of children and young people who achieve treatment response up to six months (absolute difference of 163 more per 1000 people; risk ratio (RR) 1.67, 95% confidence interval (CI) 1.21 to 2.31; I² = 0%; 3 trials, 328 participants; low‐certainty evidence). However, methotrexate compared with placebo may have little or no effect on pain as measured on an increasing scale of 0 to 100 (mean difference (MD) −1.10 points, 95% CI −9.09 to 6.88; 1 trial, 114 participants), improvement in participant global assessment of well‐being (absolute difference of 92 more per 1000 people; RR 1.23, 95% CI 0.88 to 1.72; 1 trial, 176 participants), occurrence of serious adverse events (absolute difference of 5 fewer per 1000 people; RR 0.63, 95% CI 0.04 to 8.97; 3 trials, 328 participants), and withdrawals due to adverse events (RR 3.46, 95% CI 0.60 to 19.79; 3 trials, 328 participants) up to six months. We could not estimate the absolute difference for withdrawals due to adverse events because there were no withdrawals in the placebo group. All outcomes were reported within six months of randomisation. We downgraded the certainty of the evidence to low for all outcomes due to indirectness (suboptimal dosing of methotrexate and diverse outcome measures) and imprecision (few participants and low event rates). No trials reported function or the number of participants with sustained clinically inactive disease. Serious adverse events included liver derangement, abdominal pain, and inadvertent overdose.

Methotrexate plus intra‐articular corticosteroid therapy versus intra‐articular corticosteroid therapy alone

Methotrexate plus IAGC therapy compared with IAGC therapy alone may have little or no effect on the probability of sustained clinically inactive disease or the rate of withdrawals due to adverse events up to 12 months in children and young people with the oligoarticular subtype of JIA (low‐certainty evidence). We could not calculate the absolute difference in withdrawals due to adverse events because there were no withdrawals in the control group. We are uncertain if there is any difference between the interventions in the risk of severe adverse events, because none were reported. The study did not report treatment response, function, pain, or participant global assessment of well‐being.

Methotrexate versus an alternative disease‐modifying antirheumatic drug

Methotrexate compared with leflunomide may have little or no effect on the probability of treatment response or on function, participant global assessment of well‐being, risk of serious adverse events, and rate of withdrawals due to adverse events up to four months. We downgraded the certainty of the evidence for all outcomes to low due to imprecision. The study did not report pain or sustained clinically inactive disease.

Authors' conclusions

Oral methotrexate (5 mg/m²/week to 15 mg/m²/week) compared with placebo may increase the number of children and young people achieving treatment response but may have little or no effect on pain or participant global assessment of well‐being. Oral methotrexate plus IAGC injections compared to IAGC injections alone may have little or no effect on the likelihood of sustained clinically inactive disease among children and young people with oligoarticular JIA. Similarly, methotrexate compared with leflunomide may have little or no effect on treatment response, function, and participant global assessment of well‐being. Serious adverse events due to methotrexate appear to be rare. We will update this review as new evidence becomes available to inform the living guideline.

Plain language summary

What are the benefits and risks of methotrexate for juvenile idiopathic arthritis?

Key messages

• Compared with placebo (dummy pill), methotrexate (a drug to stop or slow inflammatory arthritis) may result in more children and young people achieving an improvement in symptoms at six months, with little or no effect on pain, well‐being, risk of serious unwanted effects or the number of people stopping treatment (withdrawals) due to unwanted effects.
• Compared with no methotrexate, methotrexate may have similar benefits and harms in children and young people with oligoarticular arthritis (a type of arthritis that involves fewer than five joints) at 12 months when given together with intra‐articular (given into the joint) steroids, with little to no effect on remission or withdrawals due to unwanted effects.
• Methotrexate may have similar benefits and harms to leflunomide (an alternative anti‐inflammatory arthritis medicine) in children and young people at four months, with little to no effect on improvement in symptoms, function, well‐being, risk of serious unwanted effects, or withdrawals due to unwanted effects.

What is juvenile idiopathic arthritis?

Juvenile idiopathic arthritis is the most common rheumatic disease in childhood, where the immune system, which normally fights infection, attacks the lining of joints, making them swollen, stiff, and painful.

How is juvenile idiopathic arthritis treated?

Juvenile idiopathic arthritis is treated with a group of medications called disease‐modifying antirheumatic drugs (DMARDs), which can help prevent damage to joints and relieve pain and stiffness. Methotrexate is the most commonly used DMARD. Other treatments include nonsteroidal anti‐inflammatory drugs and steroids.

What did we want to find out?

We wanted to find out if methotrexate was better than no treatment, placebo (dummy pill), or alternative DMARDs for improving symptoms (tender and swollen joints), pain, function, and well‐being. We also wanted to know if methotrexate had any serious unwanted effects, or if people were likely to stop using methotrexate because of unwanted effects.

What did we do?

We searched for studies that investigated methotrexate compared with placebo or alternative DMARDs in children and young people with juvenile idiopathic arthritis.

What did we find?

This is an update of a Cochrane review first published in 2001. We included five trials with a total of 575 participants. Three trials compared methotrexate with placebo, one compared methotrexate plus intra‐articular (given into the joint) steroid therapy with intra‐articular steroid therapy alone, and one compared methotrexate with an alternative DMARD (leflunomide). The studies took place in Australia, Austria, Canada, Denmark, Finland, France, Germany, India, Italy, the Netherlands, New Zealand, the former Soviet Union, Spain, Switzerland, the UK, and the USA.

Methotrexate compared with placebo

Methotrexate compared with placebo may increase the likelihood of improved symptoms but may have little or no effect on pain or well‐being up to six months. The studies measured improvement in symptoms (number of tender or swollen joints and other outcomes such as pain and disability) using a composite index. At six months, 40% of participants using methotrexate and 24% of participants on placebo reported improvement in symptoms. At six months, 49% of participants on methotrexate and 40% of participants on placebo reported treatment success. The studies measured pain on a scale of 0 to 100 (0 means no pain). At six months, reported pain was 12.6 points lower in the methotrexate group and 11.5 points lower in the placebo group.

There may be little or no difference between methotrexate and placebo in the risk of serious unwanted events or the rate of withdrawal from treatment due to unwanted events up to six months. Serious unwanted events included inadvertent overdose in the methotrexate group and severe abdominal pain in the placebo group. Less than 1% of the children and young people receiving methotrexate experienced a serious unwanted event, compared to 1.4% of those receiving placebo. There were no withdrawals due to unwanted events in the placebo group and six withdrawals due to unwanted events in the methotrexate group.

No studies measured function or the number of participants with clinically inactive disease.

Other comparisons

Single studies show that methotrexate may not have any additional benefits when used along with intra‐articular glucocorticoid injection, or compared with leflunomide.

What are the limitations of the evidence?

We have little confidence in the evidence, which may underestimate the therapeutic effect of methotrexate due to suboptimal dosing, the low event rates, and the small number of participants in the trials. We are uncertain of the risk of serious unwanted events and unwanted events leading to withdrawal from treatment because of the very small number of events.

We did not compare different methods of delivering methotrexate (e.g. tablets, injection) or different doses of methotrexate.

How up to date is this evidence?

The evidence is current to 1 February 2023.

Summary of findings

Summary of findings 1. Methotrexate compared with placebo in children and young people with juvenile idiopathic arthritis.

Methotrexate compared with placebo for children and young people with juvenile idiopathic arthritis
Patient or population: children and young people with JIA Setting: outpatient Intervention: oral MTX (5–15 mg/m2/week) Comparator: placebo
Outcomes (time frame of absolute effects)	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	No. of participants (studies)	Certainty of the evidence (GRADE)	Plain language summary
	Assumed risk: placebo	Corresponding risk: oral MTX
Treatment responsea (up to 6 months)	243 per 1000	163 more per 1000 (51 more to 318 more)	RR 1.67 (1.21 to 2.31)	328 (3)	⊕ ⊕ ⊖ ⊖ Lowb	MTX may increase the number of children and young people achieving treatment response.
Sustained clinically inactive diseasec	—	—	—	—	—	No studies reported the number of participants with clinically inactive disease.
Function	—	—	—	—	—	No studies reported function.
Pain Mean improvement in pain on motion from baseline, 0–100 scale, 0 represents no pain (up to 6 months)	Mean improvement in pain with placebo was 11.5 points	MD 1.10 points better (9.08 points better to 6.88 points worse)	—	114 (1)	⊕ ⊕ ⊖ ⊖ Lowd	MTX may have little or no effect on pain.
Participant global assessment of well‐being (up to 6 months)	398 per 1000	92 more per 1000 (48 fewer to 287 more)	RR 1.23 (0.88 to 1.72)	176 (1)	⊕ ⊕ ⊖ ⊖ Lowe	MTX may have little or no effect on the number of children and young people with improved participant global assessment of well‐being.
Serious adverse events (up to 6 months)	14 per 1000	5 fewer per 1000 (13 fewer to 112 more)	RR 0.63 (0.04 to 8.97)	328 (3)	⊕ ⊕ ⊖ ⊖ Lowb	There may be little or no difference in the risk of serious adverse events between MTX and placebo.
Withdrawals due to adverse events (up to 6 months)	0 per 1000	32 per 1000	RR 3.46 (0.60 to 19.79)	328 (3)	⊕ ⊕ ⊖ ⊖ Lowb	There may be little or no difference between MTX and placebo in terms of the number of people who withdraw from treatment due to adverse events. We were unable to estimate absolute effects as there were no withdrawals in the placebo group.
*The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; JIA: juvenile idiopathic arthritis; MD: mean difference; MTX: methotrexate; RR: risk ratio; pedACR70/50/30: American College of Rheumatology paediatric 70, 30, 50 criteria; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

^a Response rate measured by pedACR70/50/30 (Giannini 1997), or as defined by study authors.
^b Downgraded once for indirectness (suboptimal methotrexate dosing in Giannini 1992 and Bharadwaj 2003, and varied definitions for treatment response in all three studies (Bharadwaj 2003; Giannini 1992; Woo 2000), with none using the ideal definition of pedACR70) and once for imprecision (low event rates).
^c Clinically active disease measured by Wallace criteria (Wallace 2011), or as defined by study author/s.
^d Downgraded once for imprecision (small number of participants) and once for indirectness (pain assessed via motion subscore instead of overall pain).
^e Downgraded once for imprecision (small number of participants) and once for indirectness (differences between the outcomes of interest and those reported: dichotomous rather than continuous data).

Summary of findings 2. Oral methotrexate plus intra‐articular glucocorticoid compared to intra‐articular glucocorticoid alone in children and young people with juvenile idiopathic arthritis.

Methotrexate plus intra‐articular glucocorticoid therapy compared to intra‐articular glucocorticoid therapy alone in children and young people with juvenile idiopathic arthritis
Patient or population: children and young people with JIA Settings: outpatient Intervention: oral MTX (15 mg/m2/week) + IAGC injection(s) Comparator: IAGC injection(s) alone
Outcomes (time frame of absolute effects)	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	No. of participants (studies)	Certainty of the evidence (GRADE)	Plain language summary
	Assumed risk: IAGC therapy alone)	Corresponding risk: oral MTX + IAGC therapy
Treatment responsea	—	—	—	—	—	Ravelli 2017 did not report the number of participants achieving treatment response.
Sustained clinically inactive diseaseb (12 months)	265 per 1000	98 more per 1000 (24 fewer to 281 more)	RR 1.37 (0.91 to 2.06)	207 (1)	⊕ ⊕ ⊖ ⊖ Lowc	MTX plus IAGC may have little or no effect on the number of children and young people with clinically inactive disease compared with IAGC alone.
Function	—	—	—	—	—	Ravelli 2017 did not report function.
Pain	—	—	—	—	—	Ravelli 2017 did not report pain.
Participant global assessment of well‐being	—	—	—	—	—	Ravelli 2017 did not report participant global assessment of well‐being.
Serious adverse events (12 months)	—	—	—	—	⊕ ⊖ ⊖ ⊖ Very lowd	We are unsure if there is any difference between MTX plus IAGC and IAGC alone in the risk of serious adverse events, because none occurred.
Withdrawals due to adverse events (12 months)	—	—	RR 8.75 (0.48 to 160.40)	207 (1)	⊕ ⊕ ⊖ ⊖ Lowc	There may be little or no difference between MTX plus IAGC versus IAGC alone in terms of the number of withdrawals due to adverse events. We could not estimate absolute effects as there were no withdrawals in the control group.
*The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IAGC: intra‐articular glucocorticoid; JIA: juvenile idiopathic arthritis; MTX: methotrexate; pedACR70/50/30: American College of Rheumatology paediatric 70, 50, 30 criteria; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

^a Response rate measured by pedACR70/50/30 (Giannini 1997), or as defined by study authors.
^b Clinically active disease measured by Wallace criteria (Wallace 2011), or as defined by study authors.
^c Downgraded once for risk of performance and detection bias and once for imprecision (small number of events).
^d Downgraded once for risk of performance and detection bias and twice for serious imprecision (no events).

Summary of findings 3. Methotrexate compared to alternative disease‐modifying antirheumatic drugs for children and young people with juvenile idiopathic arthritis.

Methotrexate compared to alternative disease‐modifying antirheumatic drugs for children and young people with juvenile idiopathic arthritis
Patient or population: children and young people with JIA Settings: outpatient Intervention: oral MTX (0.5 mg/kg/week) Comparator: oral leflunomide (20 kg bodyweight: 100 mg for 1 day, then 10 mg every other day maintenance dose; 20–40 kg bodyweight: 100 mg for 2 days, then 10 mg/day; > 40 kg bodyweight: 100 mg for 3 days, then 20 mg/day maintenance dose)
Outcomes (time frame of absolute effects)	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	No. of participants (studies)	Certainty of the evidence (GRADE)	Plain language summary
	Assumed risk: leflunomide	Corresponding risk: MTX
Treatment responsea (4 months)	426 per 1000	170 more per 1000 (30 fewer to 469 more)	RR 1.40 (0.93 to 2.10)	94 (1)	⊕ ⊕ ⊖ ⊖ Lowb	MTX compared with leflunomide may have little or no effect on the number of people achieving treatment response (pediACR70).
Sustained clinically inactive diseasec	—	—	—	—	—	Silverman 2005 did not report the number of participants with clinically inactive disease.
Function Assessed with CHAQ, 0–3 scale, lower is better (4 months)	Mean function with leflunomide was 0.59 points	MD 0.05 points worse (0.20 points better to 0.30 points worse)	—	94 (1)	⊕ ⊕ ⊖ ⊖ Lowd	MTX compared with leflunomide may have little or no effect on function.
Pain	—	—	—	—	—	Silverman 2005 did not report pain.
Participant global assessment of well‐being Assessed with 0–100 mm scale, lower is better (4 months)	Mean participant global assessment with leflunomide was 20.6 points	MD 6.10 mm better (14.28 better to 2.08 worse)	—	94 (1)	⊕ ⊕ ⊖ ⊖ Lowd	MTX compared with leflunomide may have little or no effect on participant global assessment of well‐being.
Serious adverse events (4 months)	64 per 1000	55 fewer per 1000 (63 fewer to 108 more)	RR 0.14 (0.01 to 2.69)	94 (1)	⊕ ⊕ ⊖ ⊖ Lowb	There may be little or no difference in serious adverse events between MTX and leflunomide.
Withdrawals due to adverse events (4 months)	21 per 1000	14 fewer per 1000 (20 fewer to 44 more)	RR 0.33 (0.04 to 3.09)	94 (1)	⊕ ⊕ ⊖ ⊖ Lowb	There may be little or no difference in the number of withdrawals due to adverse events between MTX and leflunomide.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CHAQ: Childhood Health Assessment Questionnaire; CI: confidence interval; JIA: juvenile idiopathic arthritis; MD: mean difference; MTX: methotrexate; pedACR70/50/30: American College of Rheumatology paediatric 70, 30, 50 criteria; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

^a Response rate measured by pedACR70/50/30 (Giannini 1997), or as defined by study authors.
^b Downgraded twice for imprecision (very few events).
^c Clinically active disease measured by Wallace criteria (Wallace 2011), or as defined by study authors.
^d Downgraded twice for imprecision (single trial with few participants, and CI including both benefit and harm).

Background

Description of the condition

Juvenile idiopathic arthritis (JIA) is the most common rheumatic disease in childhood. JIA is an umbrella term for a heterogeneous group of disorders that manifest as early onset arthritis. The International League Against Rheumatism (ILAR) classification criteria define JIA as arthritis of unknown origin that begins before the age of 16 years and persists for more than six weeks. It has several subtypes, defined by the number of joints affected, the presence of systemic symptoms, and other associated features (Petty 2004). These subtypes include oligoarticular JIA (fewer than five active joints within the first six months of disease), polyarticular JIA (five or more active joints within the first six months of disease), enthesitis‐related arthritis (affecting the entheses or the axial joints), psoriatic arthritis (inflammatory arthritis that may be associated with psoriasis), systemic JIA (associated with systemic symptoms of rash and fever), and undifferentiated arthritis (Petty 2004).

The cause of JIA is not completely understood but is probably a combination of genetic and environmental factors. The disease affects up to four per 1000 children under the age of 16 years. Girls are more commonly affected than boys in an overall ratio of approximately 3:2, but the ratio varies significantly between JIA subtypes (Thierry 2014).

Description of the intervention

The spectrum of available treatments for JIA is expanding. Commonly used treatments include nonsteroidal anti‐inflammatory drugs (NSAIDs); glucocorticoids; and disease‐modifying antirheumatic drugs (DMARDs), which can be further classified into conventional synthetic DMARDs (csDMARDs), biologic DMARDs (bDMARDs), or targeted synthetic DMARDS (tsDMARDs). The aim of contemporary treatments is to achieve disease remission for most people (Ravelli 2018).

Methotrexate has broad immunomodulatory properties and is used for a range of autoimmune diseases including inflammatory arthritis (Cronstein 2020; Hazlewood 2016; Lopez‐Olivo 2014; Wilsdon 2019). It can be administered orally or by parenteral (usually subcutaneous) injection. Methotrexate has been used since the 1970s in the treatment of rheumatoid arthritis (RA), and it is widely considered a safe and effective first‐line therapeutic option for this condition (Lopez‐Olivo 2014). One systematic review showed methotrexate to have substantial clinical benefit compared to placebo in the short‐term treatment of adults with rheumatoid arthritis (lasting 12 to 52 weeks), although the rate of discontinuation due to adverse events in people using methotrexate was 16% (Lopez‐Olivo 2014).

Methotrexate is commonly prescribed as first‐line therapy for JIA and other childhood rheumatic diseases (Rose 1990; Tiller 2018). While adverse effects such as nausea are quite common, occurring in 21% to 64% of children and young people with JIA (Bulatović 2011; Patil 2014; Ruperto 2004; van Dijkhuizen 2015), severe adverse events are rare (Graham 1992; Klein 2012).

How the intervention might work

The precise mechanism of action of methotrexate in the management of inflammatory diseases is unclear. Its clinical effect may be due to a combination of several mechanisms, such as folate antagonism, adenosine signalling, polyamine inhibition, T‐cell apoptosis, reduced cell proliferation, and modification of cytokine profiles (Braun 2009; Cronstein 2020; Friedman 2019; Tian 2007). Methotrexate can be administered orally or parenterally, although oral administration is associated with up to 15% lower bioavailability (Ferrara 2018; Jundt 1993). Routine doses of methotrexate in children and young people for rheumatologic indications range from 10 mg/m²/week to 15 mg/m²/week (Ferrara 2018). Doses higher than 15 mg/m²/week do not appear to have additional benefit (Ruperto 2004).

Why it is important to do this review

This is an update of a Cochrane review first published in 2001 and including two studies (Takken 2001). There have since been advances in the definition and measurement of disease activity states in JIA (Consolaro 2016; Wallace 2004; Wallace 2011). In addition, the Outcome Measures in Rheumatoid Arthritis Clinical Trials (OMERACT) JIA working group has endorsed a core domain set of outcome measures for clinical trials following global stakeholder input from people with JIA, caregivers, health professionals, and researchers (Morgan 2019). These outcomes include pain, joint inflammatory signs, activity limitation/physical function, patient global assessment of well‐being, and adverse events (Morgan 2019).

This review supports a living guideline for children and young people with JIA (JIA Living Guidelines). We plan to use the results of this living systematic review to inform the development of trustworthy clinical practice guidelines.

Objectives

To assess the benefits and harms of methotrexate for children and young people with juvenile idiopathic arthritis.

Methods

Criteria for considering studies for this review

Types of studies

We included parallel‐group and cross‐over randomised controlled trials (RCTs). Cluster‐randomised trials were ineligible. We applied no restrictions related to language or date of publication. We included studies published as full‐text articles or abstract only; we also considered unpublished data.

Types of participants

Eligible participants were aged 18 years or younger and had a diagnosis of JIA, according to the ILAR criteria (Petty 2004) or earlier equivalents, including the juvenile chronic arthritis (JCA) criteria from the European League Against Rheumatism (EULAR; Fantini 1977), or the juvenile rheumatoid arthritis (JRA) criteria from the American College of Rheumatology (ACR; Cassidy 1989). All JIA subtypes were eligible. We excluded studies that recruited mostly adults with JIA. Trials that included young adults (up to 22 years) with JIA were eligible if most participants (80% or more) were aged 18 years or younger.

Types of interventions

We included trials assessing methotrexate administered at any dose and via any route (oral or parenteral) in the following comparisons.

• Methotrexate versus placebo

• Methotrexate versus no treatment

• Methotrexate plus co‐intervention versus co‐intervention alone

• Methotrexate versus alternative DMARD

All alternative DMARDs for JIA were eligible as controls. These included, but were not limited to, the following treatments.

• csDMARDs (DMARDs that do not target a specific molecular structure), such as leflunomide, sulfasalazine, and hydroxychloroquine

• bDMARDs (DMARDs that are derived biologically and are designed to target specific cells or proteins involved in the inflammatory response present in JIA), such as infliximab, etanercept, adalimumab, certolizumab, golimumab, tocilizumab, sarilumab, anakinra, rilonacept, canakinumab, abatacept, and rituximab. Because these are biologically derived proteins, it is impossible to exactly reproduce their complex structure. bDMARDs can be further classified into bio‐originator DMARDs (first version) and biosimilar DMARDs (almost identical to the originator but manufactured by a different company).

• tsDMARDs (DMARDs designed to act on a specific molecular target), such as tofacitinib, baricitinib, and upadacitinib

We included studies in which participants received co‐interventions, including NSAIDs and other analgesics, provided the co‐interventions were applied similarly in all treatment groups.

We excluded studies that compared different doses of methotrexate and different routes of administration (i.e. oral versus parenteral).

Types of outcome measures

There are no standardised, core set outcome measures for JIA. OMERACT has established an updated core domain set but has not determined the optimal outcome measures for each domain (Morgan 2019).

We evaluated the following outcomes, based on the OMERACT core domains.

• Treatment response, defined by the following criteria (listed in order of preference; Giannini 1997)

  • ACR Pediatric 70 improvement criteria (improvement of at least 70% in three of six core set variables; pedACR70)

  • ACR Pediatric 50 improvement criteria (improvement of at least 50% in three of six core set variables; pedACR50)

  • ACR Pediatric 30 improvement criteria (improvement of at least 30% in three of six core set variables; pedACR30)

  • Other criteria selected by study authors

• Sustained clinically inactive disease, defined by the following criteria (listed in order of preference)

  • Wallace criteria (Wallace 2004)

  • Clinical Juvenile Arthritis Disease Activity Score (cJADAS; Consolaro 2016)

  • Other criteria selected by study authors

• Function, assessed by the Child Health Assessment Questionnaire (CHAQ) or other criteria selected by study authors (Ruperto 2001)*

• Pain, assessed by the following measures (listed in order of preference)*

  • Visual analogue scale (VAS)

  • FACES scale

  • Another scale selected by trial authors (Joos 1991)

• Participant global assessment of well‐being (reported by the child/young person or their caregiver)*

• Serious adverse events

• Withdrawals due to adverse events

*We intended to report these outcomes using continuous data, but if none were available, we also considered dichotomous data.

Timing of outcome assessment

We extracted all outcome measures at the end of the trial.

Search methods for identification of studies

The methods outlined below are specific to maintaining the review as a living systematic review in the Cochrane Library (Brooker 2019).

Electronic searches

We based our search methods on a model developed by Cochrane Musculoskeletal for a current living systematic review on rheumatoid arthritis (Hazlewood 2020). As part of our plan to develop a suite of living Cochrane Reviews for children and young people with a diagnosis of JIA to inform our living guidelines for JIA (JIA Living Guidelines), we established a 'living' JIA registry of all RCTs, irrespective of intervention.

Creating the registry involved conducting an initial search of Ovid MEDLINE, Ovid Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL; via Ovid), combining standard Cochrane search filters for 'juvenile idiopathic arthritis' and 'randomised trial'. The search was intentionally designed not to include terms for the intervention, allowing us to establish a record of all randomised trials in this condition.

The next step was searching for ongoing trials and protocols of published trials in the following trials registries.

• US National Institute of Health Ongoing Clinical Trials Registry ClinicalTrials.gov (clinicaltrials.gov)

• World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; www.who.int/ictrp/en/).

The baseline search to set up the registry of trials for JIA was conducted on 8 March 2021. Subsequent searches of CENTRAL, MEDLINE and Embase are conducted approximately every three to four months.

The search we accessed for this version of the review was conducted on 1 February 2023 in the following databases.

• CENTRAL (2022, Issue 12) via Ovid EBM reviews (searched 1 February 2023)

• MEDLINE Ovid (1946 to 30 January 2023)

• Embase Ovid (1974 to 30 January 2023)

Appendix 1 presents the search strategies for each database.

There was no restriction on language of publication.

For the regular search updates following the baseline search (approximately every three to four months), we will follow the same approach. We review our search methods and strategies approximately once a year to ensure they reflect any terminology changes in the topic area or in the database (Appendix 2).

Searching other resources

To optimise the accuracy and sensitivity of our search, we supplemented the database search by also searching for RCTs that were cited in recent and relevant systematic reviews, guidelines, or other documents published in the five years prior to the search. To obtain these documents, we conducted the following searches.

• Cochrane Reviews in the Cochrane Database of Systematic Reviews (CDSR) using the search term "juvenile arthritis" on 1 February 2023

• Non‐Cochrane systematic reviews identified through a search of the PubMed.govdatabase using the search terms "juvenile arthritis AND systematic[sb]" on 1 February 2023

• Paediatric rheumatology clinical guidelines on PubMed.govusing "methotrexate" and "juvenile arthritis" and the relevant society's abbreviated name (in parentheses below) on 1 February 2023

  • European League Against Rheumatism (EULAR)

  • American College of Rheumatology (ACR)

  • Paediatric Rheumatology International Trials Organisation (PRINTO)

  • Childhood Arthritis and Rheumatology Research Alliance (CARRA)

  • Single Hub and Access Point for Pediatric Rheumatology in Europe (SHARE)

  • British Society for Paediatric and Adolescent Rheumatology (BSPAR)

  • German Society for Pediatric Rheumatology (GKJR)

  • Australian Paediatric Rheumatology Group (APRG)

• A series of systematic reviews and network meta‐analyses of pharmacological and nonpharmacological interventions in JIA (Smith 2017)

We also searched the reference lists of the included trials and any relevant review articles retrieved from the electronic searches, to identify any other potentially relevant trials.

Data collection and analysis

Selection of studies

Initial screening of studies

We downloaded all records returned by the searches to a reference management database and removed all duplicates (Endnote Version 20.4). For living systematic reviews, the initial screening process involves an approach that combines machine learning and crowdsourcing to identify probable RCTs. First, we uploaded the search results to an RCT classifier, available in the Cochrane Register of Studies (CRS Web; Wallace 2017). The classifier uses machine learning algorithms to assign the probability of each citation being a true RCT. We considered studies with a probability of less than 1% to be non‐RCTs, and we excluded them. We planned to filter the remaining records though Cochrane Crowd (Cochrane Crowd 2019), a citizen science platform launched in 2016, before manually screening the studies (Wallace 2017). However, the search yield was manageable for the review author team, so we did not need Cochrane Crowd for this version of the review.

Further screening of studies

We imported the full‐text articles of the RCTs to Covidence for the manual screening process (Covidence). We annotated all records by type of population and by interventions included in the trial. No records were excluded at this stage. As part of our broader aim of a living registry of all JIA trials, we tagged all full‐text records in our registry in Covidence by JIA subtype and by interventions.

To assess eligibility for this methotrexate review, two review authors (WR, JT) independently screened the titles and abstracts of the tagged studies against the following questions.

• Is the intervention methotrexate? (yes/no/unclear)

• Is methotrexate compared with placebo, no treatment, or an alternative DMARD?

The same two review authors then assessed the full‐text articles of all potentially eligible studies against our other eligibility criteria.

We also identified and recorded reasons for excluding the ineligible studies.

We resolved any disagreement through discussion or by consulting a third review author, if necessary (RJ). We collated multiple reports of the same study, so that each study, rather than each report, was the unit of interest in the review.

We recorded the selection process in sufficient detail to complete a PRISMA flow diagram (PRISMA Statement 2020) and a Characteristics of excluded studies table.

Data extraction and management

Two review authors (WR, JT) independently extracted outcome data from the included studies. We extracted the number of events and number of participants per treatment group for dichotomous outcomes (treatment response, clinically inactive disease, serious adverse events, withdrawals due to adverse events), and means, standard deviations (SDs), and number of participants per treatment group for continuous outcomes (function, pain, participant global assessment of well‐being). We noted in the Characteristics of included studies table if outcome data were not reported in a usable way, or if we had transformed or estimated data from a graph.

We resolved disagreements by discussion or by involving a third review author (RJ). One review author (JT) transferred data to the Review Manager 5 file (Review Manager 2020). Two other review authors (WR, RJ) double‐checked correct data entry by comparing the data presented in the systematic review with the study reports.

We extracted the following study characteristics.

• Methods: study design, total duration of study, details of any treatment 'run‐in' period, number of study centres and location, study setting, withdrawals, and date/s of study

• Participants: number, mean age, age range, sex, disease duration, JIA subtype, diagnostic criteria used, joint count, disease duration, mean pain, mean CHAQ as a functional/participation outcome, JADAS, physician global assessment of disease activity (Filocamo 2010), participant global assessment of well‐being (Filocamo 2010), trial inclusion criteria, and trial exclusion criteria

• Interventions: dose of methotrexate, mode of administration, treatment duration, co‐intervention (name of drug, dose, duration, and frequency of co‐intervention treatment), permitted concomitant medications (including stable doses of certain analgesics, NSAIDs, and other DMARDs)

• Comparisons: comparator type (placebo, no treatment, alternative DMARD); in the case of the alternative DMARD comparator, the dose, treatment duration, treatment frequency or intervals; co‐interventions (name of drugs, dose, duration, and frequency of co‐intervention treatment); permitted concomitant medications

• Outcomes: treatment response (pedACR70, pedACR50, pedACR30, or as otherwise defined by study authors), clinically inactive disease (as per Wallace criteria, cJADAS, or as otherwise defined by study authors), mean function (CHAQ or as otherwise defined by study authors), mean pain (VAS, FACES scale, or as otherwise defined by study authors), participant global assessment of well‐being, serious adverse events, and withdrawals due to adverse events

• Characteristics of the design of the trial, outlined below in the Assessment of risk of bias in included studies section

• Notes: source of funding, notable declarations of interest of study authors, and details of any correspondence with study authors

To prevent selective inclusion of data based on the results, we defined the following rules before beginning data extraction.

• As these diseases have heterogeneous presentations, monitoring often includes a broad variety of assessments. We applied the following rules in the event of multiple outcome reporting.

  • We used the following prespecified hierarchy for the definition of clinically inactive disease.

    • Wallace criteria

    • cJADAS

    • Other criteria selected by trial authors

  • We used the following prespecified hierarchy for the definition of treatment response.

    • pedACR70

    • pedACR50

    • pedACR30

    • Other criteria selected by trial authors

• Where trials reported outcomes at multiple time points, we extracted data from the latest time point within the period of time we were interested in (up to 12 months).

• Where trialists reported both final values and change from baseline values for the same outcome, we preferentially extracted final values.

• Where trialists reported both unadjusted and adjusted values for the same outcome, we preferentially extracted adjusted values.

• We preferentially extracted mean outcome data over dichotomous outcome data for pain, function, and participant global assessment.

• For outcomes assessing benefits, we preferentially extracted data from intention‐to‐treat (ITT) analyses, or extracted the number of participants analysed at that time point if data were not available for missing participants. For outcomes assessing harms, we extracted data for participants who received the treatment to which they were randomised.

Assessment of risk of bias in included studies

Two review authors (JT, WR) independently assessed the risk of bias for each included study using the Cochrane risk of bias tool RoB 1 (Higgins 2017). We resolved any disagreements through discussion or by involving a third review author (RJ). RoB 1 covers the following domains.

• Random sequence generation (selection bias)

• Allocation concealment (selection bias)

• Blinding of participants and personnel (performance bias)

• Blinding of outcome assessor (detection bias)

• Incomplete outcome data (attrition bias)

• Selective outcome reporting (reporting bias)

• Other potential sources of bias, such as unequal application of co‐interventions across treatment groups, unit of analysis issues, unplanned interim analysis, unequal cross‐over of participants from one treatment group to another (e.g. from usual care control to the intervention), and design‐specific issues such as inadequate washout in cross‐over trials.

We assessed detection bias separately for participant‐reported outcomes (e.g. pain, treatment success, adverse events), where blinding is necessary to minimise bias, and for more objective outcomes (e.g. liver enzyme derangement), where blinding of the assessor may not influence the measurement.

We graded each study at high, low, or unclear risk in each domain, providing a justification for our judgement in the Characteristics of included studies table. We summarised the risk of bias judgements across different studies for each of the domains listed. We created a figure with the risk of bias tool to provide a summary assessment.

Where information on risk of bias related to unpublished data or correspondence with a trialist, we noted this in the Characteristics of included studies table.

When interpreting treatment effects, we considered the risk of bias for the studies that contributed to that outcome.

Measures of treatment effect

We presented simple summary data for each intervention group: number of events and number of participants per group for dichotomous outcomes; and mean, SD, and number of participants per group for continuous outcomes.

We used the Cochrane statistical software Review Manager 5.4 to analyse the data (Review Manager 2020).

For dichotomous data, we calculated risk ratios (RRs) with 95% confidence intervals (CIs).

For continuous data, we calculated the mean difference (MD) with 95% CI if different studies used the same scale to measure the outcome. When studies used different scales to measure the same conceptual outcome (e.g. function), we used the most common outcome measure as the index outcome measure, and calculated the standardised mean difference (SMD) with its 95% CI. We planned to back‐translate SMDs to a typical scale by multiplying the SMD by a typical among‐person SD (e.g. the SD of the control group at baseline from the most representative trial; Higgins 2021).

We entered continuous data presented as a scale with a consistent direction of effect across studies.

We used Review Manager 5.4 to create forest plots showing the summary data and effect estimates for each study (Review Manager 2020).

Unit of analysis issues

For trials with more than two arms, we described all study groups in the Characteristics of included studies table and included the intervention groups that met our review criteria in the analysis. When the variance of the difference between intervention and comparator was not reported, we calculated this from the variances of all trial arms. When a study included multiple relevant treatment arms (e.g. two groups treated with different doses of methotrexate), we combined groups to perform a single pairwise comparison (Higgins 2021).

Had any studies reported only differences between treatment groups, as opposed to mean effects for each group, we would have analysed the data using the generic inverse variance function.

For cross‐over trials, we extracted data from the first phase of the trial to avoid potential carry‐over effects.

Dealing with missing data

For dichotomous outcomes assessing possible benefits of treatment (e.g. sustained clinically inactive disease), we calculated the RR using the number of participants randomised to the group as the denominator. For dichotomous outcomes assessing possible harms of treatment (e.g. withdrawals due to adverse events), we calculated the RR using the number of participants randomised to the group and receiving at least one treatment as the denominator.

For continuous outcomes (e.g. pain), we calculated the MD or SMD based on the number of participants analysed at that time point. Had we been unable to determine the number of participants analysed at each time point, we would have used the number of participants randomised to each group at baseline.

We planned to contact investigators or study sponsors to verify key study characteristics and obtain missing numerical outcome data. We computed missing SDs from other statistics, such as standard errors, CIs, or P values, according to the methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions. Had we been unable to calculate SDs, we would have imputed them (e.g. from other studies in the meta‐analysis; Deeks 2021).

Assessment of heterogeneity

We assessed the clinical and methodological diversity of the included studies, in terms of participants, interventions, outcomes, and study characteristics, to determine whether a meta‐analysis was appropriate. We assessed statistical heterogeneity by visually inspecting forest plots to check for obvious differences in results between the studies, and by using the I² statistic and the Chi² test.

We interpreted I² values using the following rough thresholds, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2021).

• 0% to 40% might not be important.

• 30% to 60% may represent moderate heterogeneity.

• 50% to 90% may represent substantial heterogeneity.

• 75% to 100% represents considerable heterogeneity.

As noted in the Cochrane Handbook for Systematic Reviews of Interventions, the importance of I² depends on the magnitude and direction of effects, and the strength of evidence for heterogeneity.

We interpreted the Chi² test so that P ≤ 0.10 indicated the presence of statistical heterogeneity. Where we identified substantial heterogeneity, we reported it and investigated possible causes by following the recommendations in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2021).

Assessment of reporting biases

To assess small study effects, we planned to generate funnel plots for meta‐analyses that included at least 10 trials. If we detected asymmetry in the funnel plot, we planned to review the characteristics of the trials to assess whether the asymmetry was likely due to publication bias or other factors, such as the methodological or clinical diversity of the trials. We planned to conduct formal statistical tests to investigate funnel plot asymmetry, and follow the recommendations in Chapter 13 of the Cochrane Handbook for Systematic Reviews of Interventions (Page 2021).

To assess outcome reporting bias, we checked trial protocols against published reports. For studies published after 1 July 2005, we searched the WHO ICTRP for the a priori trial protocol (www.who.int/clinical-trials-registry-platform). If trial protocols were unavailable, we compared the outcomes reported in the methods and results sections of the trial reports.

Data synthesis

We presented results for each outcome by grouping studies with the intervention of interest and a common comparator.

Our planned comparisons were methotrexate versus placebo, methotrexate versus no treatment (in addition to any concomitant therapies), and methotrexate versus an alternative DMARD.

Within each comparison, we pooled data for each outcome in a meta‐analysis and displayed the results in forest plots. We used the random‐effects model based on the assumption that clinical diversity was likely to exist, and that different studies were estimating different intervention effects.

If interventions and outcomes were too dissimilar to be included in a meta‐analysis, we displayed the results for individual studies in a forest plot without pooling results and described the results narratively.

Subgroup analysis and investigation of heterogeneity

We planned to carry out the following subgroup analyses, subject to sufficient data, on the outcomes of treatment response and sustained clinically inactive disease for the primary comparison.

• Oral versus parenteral routes of methotrexate administration

• Lower dose (less than 15 mg/m²/week) versus higher dose (15 mg/m²/week or higher) of methotrexate

• Use of methotrexate in specific JIA subtypes (oligoarthritis, polyarthritis (rheumatoid factor negative), polyarthritis (rheumatoid factor positive), psoriatic arthritis, enthesitis related arthritis, undifferentiated arthritis)

We planned to use the formal test for subgroup interactions in Review Manager 5 and interpret the results with caution, as advised in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2021; Review Manager 2020).

Sensitivity analysis

We planned the following sensitivity analyses to investigate the robustness of two outcomes (treatment response and sustained clinically inactive disease) to potential selection and detection biases, in the primary comparison.

• Selection bias: we planned to remove trials at high or unclear risk of bias for allocation concealment from the meta‐analysis and assess whether this changed the overall treatment effect.

• Detection bias: we planned to remove the trials that reported inadequate or unclear blinding of outcome assessors from the meta‐analysis and assess whether this changed the overall treatment effect.

Summary of findings and assessment of the certainty of the evidence

We created a summary of findings table for each comparison, including the following outcomes.

• Treatment response

• Sustained clinically active disease

• Function

• Pain

• Participant global assessment of well‐being

• Serious adverse events

• Withdrawals due to adverse events

We developed the summary of findings tables using MagicApp online software (MagicApp).

Two review authors (JT, WR) independently assessed the certainty of the evidence across all studies contributing to the meta‐analysis for each outcome, using the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias), as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2021a; Schünemann 2021b). JT and WR resolved any disagreements by discussion with SW and RJ. We justified decisions to downgrade the certainty of the evidence using footnotes, and we made comments to aid the reader's understanding of the review where necessary.

Results

Description of studies

Results of the search

The search yielded 8057citations (2968 after deduplication). The RCT classifier eliminated 2706 records, and we collated the remaining 262 studies into the JIA RCT registry in Covidence. After removing clearly ineligible RCTs in the title and abstract screen, we assessed 21 full‐text articles. We eliminated 15 studies because they had an ineligible intervention (14 studies) or an ineligible population (one study). Five RCTs met the eligibility criteria for this review: two trials from the original review (Giannini 1992; Woo 2000), and three new trials (Bharadwaj 2003; Ravelli 2017; Silverman 2005). We listed one new study as awaiting classification (Burrone 2022).

Figure 1 illustrates the study selection process in a PRISMA diagram.

1.

PRISMA study flow diagram.

Included studies

The Characteristics of included studies table provides details of all included trials (methods, participants, interventions, outcomes, funding, and trial registration).

Study design, setting, and funding

Three trials evaluated methotrexate versus placebo (Bharadwaj 2003; Giannini 1992; Woo 2000), one evaluated methotrexate plus concomitant intra‐articular glucocorticoid therapy versus intra‐articular glucocorticoid therapy alone (Ravelli 2017), and one evaluated methotrexate versus leflunomide (Silverman 2005).

One RCT had a cross‐over design (Woo 2000), and one RCT included two active intervention arms (very low‐dose methotrexate and low‐dose methotrexate; Giannini 1992).

Two trials were conducted in a single country: Italy (Ravelli 2017) and India (Bharadwaj 2003). Two trials were conducted in two countries: the USA and former Soviet Union (Giannini 1992) and the UK and France (Woo 2000). One multinational trial was conducted in Australia, Austria, Canada, Denmark, Finland, France, Germany, the Netherlands, New Zealand, Spain, Switzerland, and the USA (Silverman 2005). All were conducted in the outpatient setting.

One trial received at least some funding from manufacturers (Silverman 2005). No other trials explicitly reported that they were free from manufacturer funding, and Bharadwaj 2003 did not report the source of funding.

Participants

Table 4 compares participant characteristics across trials.

1. Baseline demographic and clinical characteristics of the trial participants.

Trial (country)	Treatment groups (n)	Age and sex	JIA inclusion criteria	JIA subtype	Permitted concomitant medications	Duration of disease	Baseline medication characteristics	Clinical assessments
MTX versus placebo (3 trials)
Bharadwaj 2003 (India)	MTX (n = 14)	Age range 3–18 years 36% female	ACR criteria; 4 or more active joints unresponsive to 6 weeks' treatment with NSAIDs or steroids	14% oligoarticular onset; 64% polyarticular; 21.4% systemic onset	Glucocorticoids ≤ 0.5 mg/kg/day	Range 6–84 months	Not reported	Monthly for 6 months
	Placebo (n = 11)	Age range 2–22 years 45% female		9% oligoarticular onset; 63.6% polyarticular; 27% systemic onset		Range 6–108 months	Not reported
Giannini 1992 (USA and Russia)	Low‐dose MTX (n = 46)	Mean age 10.1 years 72% female	ACR criteria for JRA or equivalent criteria used in the Soviet Union & Eastern Europe; minimum 3 joints with active arthritis as per ACR criteria	20% systemic onset disease; 80% non‐systemic disease	≤ 2 NSAIDs, prednisolone dose ≤ 0.5 mg/kg/day (≤ 10 mg/day)	Mean 57.6 months	Low‐dose prednisone: 33% 2 NSAIDs: 11%	Monthly for 6 months
	Very low‐dose MTX (n = 40	Mean age 9.6 years 73% female		28% systemic onset disease; 72% non‐systemic disease		Mean 57.6 months	Low‐dose prednisone: 37% 2 NSAIDs: 7.5%
	Placebo (n = 41)	Mean age 10.6 years 83% female		29% systemic onset disease; 71% non‐systemic disease		Mean 69.6 months	Low‐dose prednisone: 34% 2 NSAIDs: 7.3%
Woo 2000 (UK and France)	MTX (crossover; n = 44)	Mean age 8 years 69% female	ILAR criteria for extended oligoarthritis or systemic arthritis; for children with systemic arthritis, ≥ 3 months' disease duration despite prednisolone > 5 mg/day and NSAIDs; for children with extended oligoarthritis, active polyarthritis not responding to NSAID or local steroid treatment for at least 1 year	All participants, regardless of subtype, underwent both methotrexate and placebo phase (cross‐over trial)	NSAIDs and oral prednisolone	Mean 43.7 months	Not reported	Monthly for 10 months
	Placebo (crossover; n = 44)
MTX plus intra‐articular glucocorticoid therapy versus intra‐articular glucocorticoid therapy alone (1 trial)
Ravelli 2017 (Italy)	MTX plus IAGC (n = 105)	Mean age 4.1 years; age range 2.4–8.9 years 80% female	ILAR criteria Oligoarticular JIA	100% oligo‐articular subtype JIA	Nil allowed other than the co‐intervention of IAGC injection provided; NSAIDs had to be ceased before enrolment.	Mean 6.9 months	Prednisone: 4% Previous IAGC injection: 44% NSAIDs: 11%	1, 3,6, and 12 months
	IAGC alone (n = 102)	Mean age 2.8 years; age range 1.6–6 years 72% female		100% oligo‐articular subtype JIA		Mean 6.7 months	Prednisone: 2% Previous IAGC injection: 44% NSAIDs: 9%
MTX versus alternative DMARD (1 trial)
Silverman 2005 (Australia, Austria, Canada, Denmark, Finland, France, Germany, Netherlands, New Zealand, Spain, Switzerland, USA)	MTX (n = 47)	Mean age 10.2 years 72% female	ACR criteria for JRA with a polyarticular disease course	Polyarticular JIA: 83%; pauciarticular JIA: 17%; systemic onset: 0%	NSAIDs, ≤ 0.2 mg/kg/day of prednisolone or equivalent (≤ 10 mg/day) if dose remained unchanged 2 weeks before enrolment and during study; ≤ 2 intra‐articular steroid injections (THA) allowed during study	Mean 16.44 months	Glucocorticoid use at entry: 19%	Monthly for 4 months
	Leflunomide (n = 47)	Mean age 10.1 years 75% female		Polyarticular JIA: 77%; pauciarticular JIA: 21%; systemic onset: 2%		Mean 20.28 months	Glucocorticoid use at entry: 19%

ACR: American College of Rheumatology; DMARD: disease‐modifying antirheumatic drugs; IAGC: intra‐articular glucocorticoid; ILAR: International League of Associations for Rheumatology; JIA: juvenile idiopathic arthritis; JRA: juvenile rheumatoid arthritis; MTX: methotrexate; n = number of participants; NSAID: nonsteroidal anti‐inflammatory drug; THA: triamcinolone hexacetonide.

The five trials included 575 participants, and the mean number of participants was 115 (range 31 to 226). Trial duration ranged from four to 12 months. Ravelli 2017 and Woo 2000 recruited children and young people with a diagnosis of JIA/JRA according to the ILAR classification criteria (Petty 1998); while the remaining three trials used the ACR classification criteria (Brewer 1977).

Across the five trials, most participants had a nonsystemic JIA subtype (polyarticular or oligoarticular). Two trials had no participants with systemic‐onset JIA (Ravelli 2017; Silverman 2005), and one of these two included only participants with persistent oligoarticular JIA (Ravelli 2017).

Four trials recruited children and young people up to age 18 years (Giannini 1992; Ravelli 2017; Silverman 2005; Woo 2000), while Bharadwaj 2003 did not specify an age range, and participants in the placebo group were aged up to 22 years. The mean age of participants across the trials ranged from four to 10 years. Four trials had mostly female participants, who represented between 70% and 83% of the population (Giannini 1992; Ravelli 2017; Silverman 2005; Woo 2000). The proportion of females in Bharadwaj 2003 was 36%, although most participants had nonsystemic JIA.

The mean duration of disease prior to enrolment ranged from 6.9 months to 69.6 months across four trials (Giannini 1992; Ravelli 2017; Silverman 2005; Woo 2000). Bharadwaj 2003 did not report mean duration of disease, but the range was six months to 108 months. In keeping with current clinical practice, the studies only recruited children and young people with active disease that required an escalation in therapy.

All studies reported medications at baseline, which included NSAIDs, oral prednisolone, and intra‐articular glucocorticoids.

Interventions

Table 5 compares the intervention characteristics across trials.

2. Characteristics of interventions in included trials.

Study ID	Treatment groups	Intervention	Control	Co‐interventions
Methotrexate versus placebo (3 trials)
Bharadwaj 2003	MTX (n = 14) Placebo (n = 11)	Dose: MTX 10 mg/m2/week Route: not stated Duration: 6 months	Identical placebo	Both groups received NSAIDs and (if required) corticosteroids at a dose of 0.5 mg/kg/day.
Giannini 1992	Low‐dose MTX (n = 46) Very low‐dose MTX (n = 40) Placebo (n = 41)	Dose: MTX 10 mg/m2/week (low dose) MTX 5 mg/m2/week (very low dose) Route: oral Duration: 6 months	Identical placebo	Permitted medications had to be constant 1 month prior to trial and not be changed during trial. Maximum of 2 NSAIDs; prednisolone dose ≤ 0.5 mg/kg/day (max 10 mg/day).
Woo 2000	MTX (cross‐over; n = 44) Placebo (cross‐over; n = 44)	Dose: MTX 15 mg/m2/week (option to increase to 20 mg/m2/week at 2 months) Route: oral Duration: 4‐month active/placebo treatment period, followed by 2‐month washout, then 4‐month placebo/active treatment period, followed by another 2‐month washout	Identical placebo	None reported
Methotrexate plus intra‐articular glucocorticoids versus intra‐articular glucocorticoids alone (1 trial)
Ravelli 2017	MTX plus IAGC (n = 105) IAGC alone (n = 102)	MTX Dose: 15 mg/m2/week plus folinic acid (25–50% of the MTX dose in mg, the day after MTX administration). MTX was started in all patients within 1 week of joint injection. Route: oral Duration: 12 months IAGC Dose: triamcinolone hexacetonide at a dose of 1 mg/kg (maximum 40 mg) in knees and shoulders, 0·75 mg/kg (maximum 30 mg) in ankles and elbows, and 0·25–0·5 mg/kg (maximum 20 mg) in wrists. Methylprednisolone acetate at a dose of 5 to 10 mg for small hand and foot joints and 20–40 mg for subtalar and intertarsal joints.	IAGC:triamcinolone hexacetonide at a dose of 1 mg/kg (maximum 40 mg) in knees and shoulders, 0·75 mg/kg (maximum 30 mg) in ankles and elbows, and 0·25–0·5 mg/kg (maximum 20 mg) in wrists. Methylprednisolone acetate at a dose of 5 to 10 mg for small hand and foot joints and 20–40 mg for subtalar and intertarsal joints. Most younger patients (at investigator discretion) or patients who underwent the IAGC injections in ≥ 2 joints had the procedure carried out under general anaesthesia. A non‐weight bearing period of 24 hours was prescribed after injection of glucocorticoids.	NSAID and systemic steroids not allowed
Methotrexate versus alternative disease‐modifying antirheumatic drug (1 trial)
Silverman 2005	MTX (n = 47) Leflunomide (n = 47)	Dose: MTX 0.5 mg/kg/week Route: oral Duration: 16 weeks	Leflunomide, dose dependent on bodyweight: 20 kg: 100 mg for 1 day, then 10 mg every other day maintenance dose 20–40 kg: 100 mg for 2 days followed by 10 mg/day > 40 kg: 100 mg for 3 days followed by 20 mg/day maintenance dose	All participants received at least 5 mg folate per week. Up to 2 intra‐articular steroid injections (triamcinolone hexacetonide) was allowed per participant. Study allowed NSAIDs and no more than 0.2 mg of prednisolone or equivalent per kg of body weight per day (maximum daily dose 10 mg) if dose remained unchanged 2 weeks before trial and during trial.

DMARD: disease‐modifying antirheumatic drug; MTX: methotrexate; n: number of participants; NSAID: nonsteroidal anti‐inflammatory drug.

All trials evaluated oral weekly methotrexate, but the regimen and dose varied across trials. In the three placebo‐controlled trials, the weekly methotrexate doses were 5 mg/m² and 10 mg/m² (Giannini 1992), 15 mg/m² (Woo 2000), and 10 mg/m² (Bharadwaj 2003).

One trial compared 0.5 mg/kg methotrexate to oral leflunomide. The leflunomide dose varied according to the participant's body weight. Participants under 20 kg were given 100 mg on the first day followed by 10 mg every other day; participants between 20 kg and 40 kg were given 100 mg daily for two days followed by 10 mg daily; and participants over 40 kg were given 100 mg daily for three days followed by 20 mg daily (Silverman 2005).

The final trial evaluated methotrexate 15 mg/m²/week plus intra‐articular glucocorticoids versus intra‐articular glucocorticoids alone (Ravelli 2017). The intra‐articular glucocorticoid was administered as a single injection or single set of injections at baseline: triamcinolone hexacetonide 1 mg/kg (maximum 40 mg) in knees and shoulders, triamcinolone hexacetonide 0.75 mg/kg (maximum 30 mg) in ankles and elbows, triamcinolone hexacetonide 0.25 mg/kg to 0.5 mg/kg (maximum 20 mg) in wrists, triamcinolone hexacetonide 20 mg to 40 mg for subtalar and intertarsal joints, and methylprednisolone acetate 5 mg to10 mg for small hand and foot joints (Table 5).

Four trials allowed NSAIDs, oral glucocorticoids, or both (Bharadwaj 2003; Giannini 1992; Silverman 2005; Woo 2000). Participants in Ravelli 2017 had to have ceased NSAIDs prior to enrolment.

Outcomes

Table 6 compares the reported outcome measures across trials.

3. Outcome Reporting Bias In Trials (ORBIT) matrix.

Study ID	Treatment response	Sustained clinically inactive disease	Function	Pain	Participant global assessment of well‐being	Serious adverse events	Withdrawal due to adverse events
Methotrexate versus placebo (3 trials)
Bharadwaj 2003	Yes: 50% improvement in ILAR criteria	No	Measured but not reported (Steinbroker's functional grade)	Measured (by VAS) but no measures of variances reported.	Measured and described as patient global improvement (tool not clarified in study methods), but no measures of variances reported.	Yes	Yes
Giannini 1992	Yes: composite index (index of improvement)	No	No	Yes: pain severity score (severity of pain on motion) measured as median change from baseline.	No	Yes	Yes
Woo 2000	Yes: clinically significant improvement of 30% in ≥ 3 variables (partial Giannini criteria)	No	No	No	Parent global assessment of disease activity measured as continuous outcome and reported as dichotomous outcome (proportion reporting improvement).	Yes	Yes
Methotrexate plus intra‐articular glucocorticoids versus intra‐articular glucocorticoids (1 trial)
Ravelli 2017	No	Yes	No	No	No	Yes	Yes
Methotrexate versus alternative disease‐modifying antirheumatic drug (1 trial)
Silverman 2005	Yes (pedACR70)	No	Yes (CHAQ Disability Index)	No	Yes	Yes	Yes

CHAQ: Childhood Health Assessment Questionnaire; ILAR: International League of Associations for Rheumatology classification criteria; pedACR70: American College of Rheumatology pediatric 70 improvement criteria; VAS: visual analogue scale.

Treatment response

Silverman 2005 measured the proportion of participants with treatment response using the pedACR70 (Giannini 1997), Bharadwaj 2003 measured 50% improvement using modified WHO/ILAR criteria for assessment of response in rheumatoid arthritis, Giannini 1992 measured 25% improvement using a composite index developed for a prior study, and Woo 2000 measured 30% improvement using five of the six Giannini criteria (Giannini 1997).

One trial did not measure treatment response (Ravelli 2017).

Sustained clinically inactive disease

Only Ravelli 2017 reported the proportion of participants with sustained clinically inactive disease, using the Wallace criteria (Wallace 2004).

Function

Only Silverman 2005 reported function, measured using the CHAQ, which is a scale from 0 to 3 where 0 represents optimal participant function (Ruperto 2001).

Pain

Giannini 1992 reported pain as mean improvement of pain on motion from baseline. We analysed these data because no other trial provided usable data for this outcome. Bharadwaj 2003 measured pain but only reported the results in terms of the statistical significance of the difference between groups, without providing the outcome data.

Participant global assessment of well‐being

Three trials reported participant global assessment of well‐being (Bharadwaj 2003; Silverman 2005; Woo 2000). Woo 2000 reported participant global assessment as a dichotomous outcome, as a proportion of participant's global assessment of improvement, which we substituted for this outcome due to the absence of alternative data in the three placebo‐controlled trials (Bharadwaj 2003; Giannini 1992; Woo 2000). Bharadwaj 2003 reported the results in terms of the statistical significance of the difference between groups without providing the outcome data. Silverman 2005, which compared methotrexate with an alternative DMARD (leflunomide), reported a continuous outcome of participant global assessment of well‐being on a VAS from 0 (disease inactive) to 100 (maximal disease activity).

Serious adverse events

All trials recorded serious adverse events.

Withdrawals due to adverse events

All trials recorded withdrawals due to adverse events.

Excluded studies

We excluded 15 studies after full‐text assessment (see Characteristics of excluded studies). One study recruited mostly adults (Romero 2017); four compared different doses or modes of administration of methotrexate (Dupuis 1995; Foell 2010; Ruperto 2004; Verkaaik 2011); six included methotrexate in both treatment arms to evaluate the effects of adding another medicine to methotrexate alone (Gao 2003; Hissink 2019; NL5742; Rezaieyazdi 2023; Tynjälä 2011; Wallace 2012); and four studies with titles that appeared relevant did not use methotrexate in either treatment arm (EUDRACT 2018‐001571‐21; Forster 2000; Kvien 1985; NCT04614311).

Studies awaiting classification

One study is awaiting classification because the protocol provides insufficient information to determine eligibility for this review (Burrone 2022; Characteristics of studies awaiting classification). Burrone 2022 has two intervention arms: the 'Step Down' arm will include methotrexate in the treatment regimen from onset in addition to intra‐articular glucocorticoid injection or etanercept, while participants in the 'Step Up' arm with less severe disease will have intra‐articular glucocorticoid injections only, before stepping up to methotrexate only if there is ongoing disease after glucocorticoid injections.

Risk of bias in included studies

Figure 2 summarises the risk of bias assessment results, while the Characteristics of included studies table includes justifications for all judgements.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

We judged two trials at low risk of bias across all domains (Giannini 1992; Silverman 2005), while Woo 2000 provided insufficient information for us to judge one domain. The other two trials were at high risk of bias in at least one domain (Bharadwaj 2003; Ravelli 2017).

Allocation

Four trials were at low risk of selection bias (Giannini 1992; Ravelli 2017; Silverman 2005; Woo 2000). Bharadwaj 2003 provided no details of randomisation and allocation concealment methods (unclear risk of bias).

Blinding

All three placebo‐controlled trials were at low risk of performance and detection bias (Bharadwaj 2003; Giannini 1992; Woo 2000). We also considered Silverman 2005 at low risk of performance and detection bias, as it used the double‐dummy method to blind participants to treatment allocation. We judged Ravelli 2017 at high risk of performance and detection bias due to lack of participant blinding.

Incomplete outcome data

As all trials had low dropout rates and similar dropout rates between study arms, they were at low risk of attrition bias (Bharadwaj 2003; Giannini 1992; Ravelli 2017; Silverman 2005; Woo 2000).

Selective reporting

We judged two trials at low risk of reporting bias (Giannini 1992; Silverman 2005).

Bharadwaj 2003 was unregistered and had no published protocol available. In addition, it used unvalidated WHO/ILAR criteria for 50% treatment response for rheumatoid arthritis rather than a validated JIA treatment response measure, and results for some outcomes listed in the methods section were not reported. For these reasons, we judged the study at high risk of reporting bias.

We judged Ravelli 2017 at unclear risk of reporting bias because it only reported adverse events for the intervention arm, and it was unclear if any adverse events occurred in the control arm. We found no published protocol or core outcome sets for Woo 2000, which we also considered at unclear risk of bias in this domain.

Other potential sources of bias

We identified no other sources of bias for any trials.

Effects of interventions

See: Table 1; Table 2; Table 3

We created summary of findings tables for methotrexate versus placebo (Table 1), methotrexate plus intra‐articular glucocorticoid therapy versus intra‐articular glucocorticoid therapy alone (Table 2), and methotrexate versus leflunomide (Table 3).

Methotrexate versus placebo

We analysed data from all three placebo‐controlled trials (Bharadwaj 2003; Giannini 1992; Woo 2000). Giannini 1997 had three arms that met our eligibility criteria; we merged the two intervention arms into a single arm for analysis. We extracted data from the first phase of the cross‐over trial (Woo 2000).

Treatment response

All three trials reported treatment response (Bharadwaj 2003; Giannini 1992; Woo 2000). Oral methotrexate (5 mg/m²/week to 15 mg/m²/week) may increase the number of children and young people achieving treatment response up to six months (RR 1.67, 95% CI 1.21 to 2.31; 328 participants; low‐certainty evidence; Analysis 1.1). The absolute effect estimate in the methotrexate group was 406 per 1000 compared with 243 per 1000 in the placebo group (between‐group difference of 163 more per 1000, 95% CI 51 more to 318 more). We downgraded the certainty of the evidence once for indirectness and once for imprecision.

1.1. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 1: Treatment response

We were unable to perform our predefined subgroup analyses because all studies used oral methotrexate at a dose of up to 15 mg, and no trials reported JIA subtypes.

Bharadwaj 2003 was at unclear risk of bias for allocation concealment. When we removed this trial in a sensitivity analysis, there was little difference in the effect estimate (RR 1.65, 95% CI 1.04, 2.62). We could not perform a sensitivity analysis based on blinding of outcome assessors, as all studies were at low risk of detection bias.

Sustained clinically inactive disease

No trials reported the number of participants with sustained clinically inactive disease.

Function

No trials measured function.

Pain

Only Giannini 1992 measured pain, using a subscale of 0 to 100 (with higher scores indicating more pain). Methotrexate compared with placebo may have little or no effect on pain. At six months, mean change inpain was 12.6 points lower with methotrexate and 11.5 points lower with placebo (MD −1.10 points, 95% CI −9.08 to 6.88; 114 participants; low‐certainty evidence; Analysis 1.4). We downgraded the certainty of the evidence once for indirectness and once for imprecision.

1.4. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 4: Pain (visual analogue scale)

Participant global assessment of well‐being

Only Woo 2000 reported participant global assessment of well‐being, using dichotomous data (significant improvement, yes/no). Methotrexate compared with placebo may have little or no effect on participant global assessment of well‐being (RR 1.23, 95% CI 0.88 to 1.72; 176 participants; low‐certainty evidence; Analysis 1.5). The absolute effect estimate was 490 per 1000 people with methotrexate versus 398 per 1000 people with placebo (between‐group difference of 92 more per 1000 people, 95% CI 48 fewer to 287 more). We downgraded the certainty of the evidence once for indirectness and once for imprecision.

1.5. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 5: Participant global assessment of well‐being

Serious adverse events

All three trials reported serious adverse events (Bharadwaj 2003; Giannini 1992; Woo 2000). There may be little or no difference in the risk of serious adverse events between methotrexate and placebo (RR 0.63, 95% CI 0.04 to 8.97; 328 participants; low‐certainty evidence; Analysis 1.6). The absolute effect estimate was 9 per 1000 with methotrexate versus 14 per 1000 with placebo (between group difference of 5 fewer, 95% CI 13 fewer to 112 more). We downgraded the certainty of the evidence once for indirectness and once for imprecision.

1.6. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 6: Serious adverse events

Serious adverse events included an inadvertent overdose in the methotrexate group (Bharadwaj 2003), and one case of severe abdominal pain in the placebo group (Giannini 1992).

Withdrawals due to adverse events

All three trials recorded withdrawals due to adverse events (Bharadwaj 2003; Giannini 1992; Woo 2000). There may be little or no difference between methotrexate and placebo in the number of children and young people withdrawing from treatment due to adverse events (RR 3.46, 95% CI 0.60 to 19.79; 328 participants; low‐certainty evidence; Analysis 1.7). We were unable to estimate absolute effects because there were no withdrawals due to adverse events in the placebo group. We downgraded the certainty of the evidence once for indirectness and once for imprecision.

1.7. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 7: Withdrawals due to adverse events

Adverse events that resulted in withdrawal included one inadvertent overdose (Bharadwaj 2003), two cases of liver enzyme derangement (Giannini 1992; Woo 2000), one case of intercurrent hepatitis A infection (Woo 2000), one case of persistent haematuria (Giannini 1992), and one case of persistent skin rash (Giannini 1992).

Methotrexate versus no treatment

We identified no studies that evaluated methotrexate versus no treatment.

Methotrexate plus intra‐articular glucocorticoid therapy versus intra‐articular glucocorticoid therapy alone

Only Ravelli 2017 provided data for this analysis.

Treatment response

Ravelli 2017 did not report treatment response.

Sustained clinically inactive disease

Ravelli 2017 reported sustained clinically inactive disease at 12 months. Methotrexate plus intra‐articular glucocorticoid therapy compared with intra‐articular glucocorticoid therapy alone may have little or no effect on the number of children and young people with sustained clinically inactive disease (RR 1.37, 95% CI 0.91 to 2.06; 207 participants, low‐certainty evidence; Analysis 2.2). The absolute effect estimate was 363 per 1000 people with methotrexate plus intra‐articular glucocorticoid therapy versus 265 per 1000 people for intra‐articular glucocorticoid therapy alone (between‐group difference of 98 more per 1000, 95% CI 24 fewer to 281 more). We downgraded the certainty of the evidence once for risk of bias and once for imprecision.

2.2. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 2: Sustained clinically inactive disease (Wallace criteria or as defined by study authors)

Function

Ravelli 2017 did not report function.

Pain

Ravelli 2017 did not report pain.

Participant global assessment of well‐being

Ravelli 2017 did not report participant global assessment of well‐being.

Serious adverse events

Ravelli 2017 reported serious adverse events at 12 months. We are uncertain if there is any difference in the risk of serious adverse events with methotrexate plus intra‐articular glucocorticoid therapy versus intra‐articular glucocorticoid therapy alone, as no participants in either group reported a serious adverse event (207 participants; very low‐certainty evidence). We downgraded the certainty of the evidence once for risk of bias and twice for imprecision.

Withdrawals due to adverse events

Ravelli 2017 reported withdrawals due to adverse events at 12 months. Methotrexate plus intra‐articular glucocorticoid therapy compared to intra‐articular glucocorticoid therapy alone may have little or no effect on the number of children and young people withdrawing from treatment due to adverse events (RR 8.75, 95% CI 0.48 to 160.40; 207 participants; low‐certainty evidence; Analysis 2.7). We were unable to estimate absolute effects because there were no withdrawals due to adverse events in the control group. We downgraded the certainty of the evidence once for risk of bias and once for imprecision.

2.7. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 7: Withdrawals due to adverse events

Adverse events that led to withdrawal included one case of increased liver enzyme derangement, one case of gastrointestinal discomfort, and two cases of iridocyclitis (one leading to a change in medication).

Methotrexate versus alternative disease‐modifying antirheumatic drugs

Only Silverman 2005 provided data for this comparison.

Treatment response

Silverman 2005 reported treatment response (pediACR70) at four months. Methotrexate compared with leflunomide may have little or no effect on the number of children and young people achieving treatment response (RR 1.40, 95% 0.93 to 2.10; 94 participants; low‐certainty evidence; Analysis 3.1). The absolute effect estimate was 596 per 1000 people with methotrexate versus 426 per 1000 people with leflunomide (between‐group difference of 170 more, 95% CI 30 fewer to 469 more). We downgraded the certainty of the evidence twice for imprecision.

3.1. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 1: Treatment response (American College of Rheumatology criteria or as defined by study authors)

Sustained clinically inactive disease

Silverman 2005 did not report the number of participants with sustained clinically inactive disease.

Function

Silverman 2005 assessed function using the CHAQ (scale of 0 to 3, lower is better) at four months. Methotrexate compared with leflunomide may have little or no effect on function (MD 0.05 points, 95% CI −0.20 to 0.30; 94 participants; low‐certainty evidence; Analysis 3.3). Mean function was 0.64 in the methotrexate group versus 0.59 in the leflunomide group. We downgraded the certainty of the evidence twice for imprecision.

3.3. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 3: Function (Childhood Health Assessment Questionnaire or equivalent)

Pain

Silverman 2005 did not report pain.

Participant global assessment of well‐being

Silverman 2005 reported participant global assessment of well‐being (scale of 0 mm to 100 mm, lower is better) at four months. Methotrexate compared with leflunomide may have little or no effect on participant global assessment of well‐being (MD −6.10 mm, 95% CI −14.28 to 2.08; 94 participants; low‐certainty evidence; Analysis 3.5). The mean score was 14.5 mm in the methotrexate group versus 20.6 mm in the leflunomide group. We downgraded the certainty of the evidence twice for imprecision.

3.5. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 5: Participant global assessment of well‐being

Serious adverse events

Silverman 2005 reported serious adverse events at four months. There may be little or no difference in the risk of serious adverse events between methotrexate and leflunomide (RR 0.14, 95% 0.01 to 2.69; 94 participants, low‐certainty evidence; Analysis 3.6). There was an overall low event rate in both groups, 9 per 1000 with methotrexate versus 64 per 1000 with leflunomide (between group difference of 55 fewer per 1000, 95% CI 63 fewer to 108 more). We downgraded the certainty of the evidence twice for imprecision.

3.6. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 6: Serious adverse events

No serious adverse events were reported in the methotrexate group. Serious adverse events in the leflunomide group included one case of salmonellosis, one case of liver enzyme derangement, and one case of parapsoriasis.

Withdrawals due to adverse events

Silverman 2005 reported withdrawals due to adverse events at four months. There may be little or no difference in the rate of withdrawals due to adverse events between methotrexate and leflunomide (RR 0.33, CI 95% 0.04 to 3.09; 94 participants; low‐certainty evidence; Analysis 3.7). The withdrawal rate was 21 per 1000 with methotrexate compared to 7 per 1000 with leflunomide (between‐group difference of 14 fewer per 1000, 95% CI 20 fewer to 44 more). We downgraded the certainty of the evidence twice for imprecision.

3.7. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 7: Withdrawals due to adverse events

One case of liver enzyme derangement led to treatment withdrawal in the methotrexate group. Adverse events that led to withdrawal in the leflunomide group included one case of liver enzyme derangement, one case of parapsoriasis, and one case of Crohn's disease (deemed unrelated to the study medication).

Discussion

Summary of main results

This update included three new RCTs, bringing the total to five (575 participants). Three trials compared oral methotrexate with placebo, one compared methotrexate plus intra‐articular glucocorticoid therapy with intra‐articular glucocorticoid therapy alone, and one compared methotrexate with leflunomide. Doses of methotrexate ranged from 5 mg/m²/week to 15 mg/m²/week in four trials, and participants in the methotrexate group of the remaining trial received 0.5 mg/kg/week. The number of participants varied from 31 to 226. Two placebo‐controlled trials and the leflunomide‐controlled trial were adequately randomised and blinded, and likely not susceptible to important biases. One placebo‐controlled trial may have been susceptible to selection bias due to lack of adequate reporting of randomisation methods. The trial investigating the addition of methotrexate to intra‐articular glucocorticoid therapy was susceptible to performance and detection biases. One study reported pharmaceutical funding (Silverman 2005).

Overall, we found little to no effect of methotrexate for most outcomes in all three comparisons.

Methotrexate (5 mg/m²/week to 15 mg/m²/week) compared to placebo may increase the number of children and young people achieving treatment response, but may have little or no effect on pain, participant global assessment of well‐being, risk of serious adverse events, and rate of withdrawals due to adverse events up to six months (Table 1). The certainty of the evidence was low for all outcomes. No placebo‐controlled studies measured function or the proportion of children and young people with sustained clinically inactive disease.

Compared with intra‐articular glucocorticoid injection alone, the addition of methotrexate (15 mg/m²/week) may have little or no effect on the proportion of children and young people with sustained clinically inactive disease and on the rate of withdrawal due to adverse events up to 12 months (Table 2). The evidence came from a single study and was of low certainty for both outcomes. We are unsure if there is any difference between the interventions in the risk of serious adverse events, because no serious adverse events were reported (very low‐certainty evidence). The study did not report function, pain, participant global assessment of well‐being, or the number of participants who achieved treatment response.

Methotrexate (0.5 mg/kg/week) compared with leflunomide may have little to no effect on treatment response, function, participant global assessment of well‐being, risk of serious adverse events, or rate of withdrawal due to adverse events up to four months (Table 3). The evidence came from a single study and was of low certainty for all outcomes. The study did not report pain or the number of participants with sustained clinically inactive disease.

Overall completeness and applicability of evidence

Participants in the included studies appeared to be representative of the general population of children and young people with JIA seen in clinical care, in terms of proportion of participants with nonsystemic JIA, proportion of female participants, and mean age. The trials were conducted in various countries, mostly in Europe, North America, and Australasia. Bharadwaj 2003, which was conducted in India, had a higher proportion of males with enthesitis‐related JIA compared with multiethnic epidemiological studies (Saurenmann 2007). In addition, two trials were published prior to the widespread use of bDMARDs (Giannini 1992; Woo 2000), such as the TNF‐alpha inhibitors in JIA (Lovell 2000).

The methotrexate treatment period in the trials ranged from four months (Silverman 2005; Woo 2000) to 12 months (Ravelli 2017), and the dose of methotrexate ranged from 5 mg/m²/week (Giannini 1992) to 15 mg/m²/week (Woo 2000), with one trial using a dose based on bodyweight (0.5 mg/kg/week) rather than body surface area (Silverman 2005).

According to contemporary guidelines and usual care, the ideal therapeutic dosing of methotrexate in JIA is up to 15 mg/m² (Beukelman 2011; Ferrara 2018). This calls into question the applicability of some of the included trials to current practice. Similarly, a follow‐up time of four months in some trials may have been too short to identify important benefits (Silverman 2005; Woo 2000). Methotrexate polyglutamate concentrations, thought to be important in the therapeutic effect of methotrexate, may take over 12 months to reach steady state concentrations in some tissues (Korell 2014).

Other reasons why the included trials may have underestimated the treatment effect of methotrexate include lack of dose escalation and inclusion of participants with very longstanding disease (up to 14.4 years).

Studies have used a wide range of outcome measures to assess disease activity in JIA (Consolaro 2016). With the development of multiple, highly effective treatments for JIA and a move towards treat‐to‐target strategies, outcome measures have also evolved (Hinze 2015; Ravelli 2018). Outcomes such as radiographic joint damage are now far less relevant, while ambitious targets such as clinically inactive disease or remission are now considered very important outcomes in JIA trials. The outcomes we considered to be most important were based on Cochrane Musculoskeletal JIA proposed outcomes and are consistent with the core domain set outlined by OMERACT (Morgan 2019). However, the included trials were conducted prior to the establishment of the JIA working group for OMERACT, so their reporting of OMERACT‐specified domains was limited. For example, while we could include the data from all three placebo‐controlled trials in our analysis of treatment response, each trial measured this outcome differently.

There were several other limitations across the included trials. Bharadwaj 2003 had a small sample size, used a lower dose of methotrexate than recommended in current guidelines, and recruited people aged up to 22 years without specifying how many participants were aged 18 and older. Giannini 1992 included outcomes that are no longer used routinely for either research or clinical purposes (e.g. articular severity score). In Ravelli 2017, the ITT analysis did not take into account a high proportion of participants who had remission in injected joints but were considered as having treatment failure based upon new onset of disease in previously unaffected joints before the 12‐month evaluation. Ravelli 2017 also had a high number of early withdrawals in the intervention group (13 within two months), and neither participants nor assessors were blinded to treatment allocation. Woo 2000 may have had insufficient power to detect an important between‐group difference in outcomes. While the cross‐over design attempts to offset this, it also presupposes that disease activity would return to baseline after a washout period, which may not be true. The other limitation of note in Woo 2000 was that the investigators did not use Giannini's full core set of criteria, as they measured outcomes before the publication of functional assessment scores as part of the core set; thus, Woo 2000 did not adequately measure the sixth core measure (functional assessment). Silverman 2005 may have had insufficient power to detect important differences in outcomes between methotrexate and leflunomide.

In summary, although studies included children and young people representative of those seen in current practice, the interventions assessed in most trials do not reflect current management of JIA. Trials largely assessed subtherapeutic doses of methotrexate for only a few months, without dose escalation. Further, two included trials pre‐date the widespread use of bDMARDs and tsDMARDs (Giannini 1992; Woo 2000), which are often incorporated into treatment. Thus, the effects seen in the trials may be smaller than those seen in practice. In addition, the aim of contemporary treatment is disease remission, but most trials did not measure clinically inactive disease or remission. Furthermore, the sample sizes and study durations were insufficient to assess the risk of adverse events or serious adverse events, so we were unable to adequately measure the relative harms of methotrexate in this population.

Quality of the evidence

The certainty of evidence was low for all outcomes except one (serious adverse events in the comparison of methotrexate plus intra‐articular glucocorticoid therapy versus intra‐articular glucocorticoid therapy alone), which had very low‐certainty evidence.

Evidence from the three placebo‐controlled trials was of low certainty for treatment response, pain, participant global assessment of well‐being, serious adverse events, and withdrawals due to adverse events (Bharadwaj 2003; Giannini 1992; Woo 2000). We downgraded all outcomes once for indirectness and once for imprecision. The indirectness was due to suboptimal dosing of methotrexate in Bharadwaj 2003 and Giannini 1992, and dissimilar or unexpected outcome measures: each trial used a different measure of treatment response (and none used pedACR70), Woo 2000 used a dichotomous measure for participant global assessment of well‐being rather than a continuous scale (and described the outcome as a "parent global assessment of disease activity" instead of well‐being), and Giannini 1992 used a pain subscale instead of measuring overall pain. The imprecision was due to low numbers of participants and events. As all outcomes were self‐reported, we considered it particularly important that there was a low risk of detection and performance biases, as these biases tend to overestimate treatment effects. As all trials for this comparison were at low risk of these biases, we did not downgrade the evidence for risk of bias concerns. We also judged that the possible selection bias present in Bharadwaj 2003 did not impact the outcomes, as evidenced by sensitivity analysis. Bharadwaj 2003 also used WHO/ILAR criteria for treatment response (unvalidated for JIA), but inspection of the forest plot suggested that this decision did not impact the overall findings for this outcome.

Evidence from a single trial comparing methotrexate plus intra‐articular glucocorticoid therapy with intra‐articular glucocorticoid therapy alone was of low certainty for sustained clinically inactive disease and withdrawal due to adverse events, and very low certainty for serious adverse events (Ravelli 2017). We downgraded the certainty of the evidence for all three outcomes once due to risk of performance and detection bias. For sustained clinically inactive disease and withdrawal due to adverse events, we downgraded by another level for imprecision (small number of events), and for serious adverse events, we downgraded twice for imprecision (no events).

Evidence from a single trial comparing methotrexate with leflunomide was of low certainty for treatment response, function, participant global assessment of well‐being, serious adverse events, and withdrawal due to adverse events (Silverman 2005). We downgraded twice for imprecision (few events or few participants) for all outcomes.

Potential biases in the review process

We established a comprehensive registry of trials for all interventions for JIA to inform this review, and we believe we are unlikely to have missed relevant trials. Up to three review authors were involved in the screening and selection of studies, the risk of bias assessment, and the GRADE assessment, ensuring rigour in the conduct of the review.

There were too few studies to formally assess the presence of publication bias. We identified only one ongoing trial, which we have listed as awaiting classification (Burrone 2022). The inclusion of this trial is unlikely to impact our results.

Agreements and disagreements with other studies or reviews

The inclusion of three new trials in this update has not changed the conclusions of the review, which was first published in 2001 (Takken 2001).

We identified no other systematic reviews on the use of methotrexate in JIA.

Authors' conclusions

Implications for practice.

The findings of this review demonstrate that in children and young people with juvenile idiopathic arthritis (JIA), methotrexate compared with placebo may increase the likelihood of achieving treatment response, but may have little or no effect on pain or participant global assessment of well‐being. Methotrexate may offer no additional benefit for children with oligoarticular JIA who receive intra‐articular glucocorticoid therapy, possibly owing to the prolonged duration of effect of intra‐articular triamcinolone hexacetonide in JIA (Zulian 2004). Methotrexate and leflunomide may have similar benefits and harms in children and young people with JIA.

However, this review may underestimate the therapeutic effect of methotrexate due to suboptimal dosing. In children with a contraindication or intolerance to methotrexate, leflunomide may be a viable alternative given comparable head‐to‐head benefits and harms. Serious adverse events and withdrawals due to adverse events appear to be rare in children and young people who receive methotrexate treatment.

Implications for research.

Although methotrexate is well established in standard clinical practice as the mainstay first‐line disease‐modifying antirheumatic drug (DMARD) for children and young people with JIA, this seems to be based on low‐ and very low‐certainty evidence. However, further placebo‐controlled trials are unlikely to be acceptable or feasible, as contemporary management of JIA aims to achieve clinical remission of the disease, and it would be clinically unacceptable to enrol people in a trial without active treatment.

Further research is needed to better understand the role of methotrexate in the era of biological DMARDs and targeted synthetic DMARDs. Future trials should include methotrexate at current recommended dosing and duration and use contemporary outcome measures with ambitious targets (including achievement of inactive disease) and patient‐focussed outcomes as outlined in the OMERACT (Outcome Measures in Rheumatology Clinical Trials) updated JIA core domain set (Morgan 2019). They should also adhere to CONSORT (Consolidated Standards of Reporting Trials) standards with respect to reporting baseline characteristics, dropouts, blinding of study participants and study personnel, randomisation, and treatment allocation.

What's new

Date	Event	Description
9 February 2024	New citation required but conclusions have not changed	We included a total of five trials (575 participants) in this review: three new trials (410 participants) and two trials from the previous review (165 participants). The conclusion that methotrexate has a small benefit in children with JIA has not changed since the previous version of the review (Takken 2001).
9 February 2024	New search has been performed	Review updated by a new author team except for one author from the previous team (Tim Takken). The search was conducted up to 1 February 2023. The earlier review included two trials (Giannini 1992; Woo 2000). An additional three new trials were included in this update (Bharadwaj 2003; Ravelli 2017; Silverman 2005).

History

Review first published: Issue 3, 2001

Date	Event	Description
21 December 2007	New citation required and conclusions have changed	Substantive amendment

Appendices

Appendix 1. Search strategies

Databases

Search date: 1 February 2023

EBM Reviews ‐ Cochrane Central Register of Controlled Trials <December 2022>

Embase Classic+Embase <1947 to 2023 January 30>

Ovid MEDLINE(R) and Epub Ahead of Print, In‐Process, In‐Data‐Review & Other Non‐Indexed Citations, Daily and Versions <1946 to January 30, 2023>

1 exp Arthritis, Juvenile Rheumatoid/ 38358

2 JIA.tw. 17175

3 (juvenile adj2 arthritis).tw. 32589

4 or/1‐3 46851

5 enthesitis.tw. 8537

6 Arthritis, Psoriatic/ 28172

7 oligoarthritis.tw. 3368

8 or/5‐7 36826

9 (child* or adolescent* or infan*).tw. 5136034

10 8 and 9 3020

11 4 or 10 47632

12 randomized controlled trial.pt. 585513

13 controlled clinical trial.pt. 95170

14 randomized.ab. 2103523

15 placebo.ab. 906151

16 clinical trials as topic.sh. 234095

17 randomly.ab. 1230757

18 trial.ti. 1057532

19 or/12‐18 4299265

20 exp animals/ not humans.sh. 37089152

21 19 not 20 2655366

22 11 and 21 1574

23 juvenile rheumatoid arthritis/ 36556

24 JIA.tw. 17175

25 (juvenile adj2 arthritis).tw. 32589

26 or/23‐25 46352

27 enthesitis.tw. 8537

28 psoriatic arthritis/ 38522

29 oligoarthritis.tw. 3368

30 or/27‐29 46123

31 (child* or adolescent* or infan*).tw. 5136034

32 limit 30 to (infant or child or preschool child <1 to 6 years> or school child <7 to 12 years> or adolescent <13 to 17 years>) [Limit not valid in Ovid MEDLINE(R),Ovid MEDLINE(R) Daily Update,Ovid MEDLINE(R) PubMed not MEDLINE,Ovid MEDLINE(R) In‐Process,Ovid MEDLINE(R) Publisher,CCTR; records were retained] 5130

33 31 or 32 5138612

34 30 and 33 5893

35 26 or 34 49255

36 Randomized controlled trial/ 1347411

37 Controlled clinical study/ 468194

38 random$.ti,ab. 4417575

39 randomization/ 204817

40 intermethod comparison/ 289732

41 placebo.ti,ab. 942788

42 (compare or compared or comparison).ti. 1217106

43 ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. 4757698

44 (open adj label).ti,ab. 228164

45 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. 759964

46 double blind procedure/ 207341

47 parallel group$1.ti,ab. 92575

48 (crossover or cross over).ti,ab. 317058

49 ((assign$ or match or matched or allocation) adj5 (alternate or group$1 or intervention$1 or patient$1 or subject$1 or participant$1)).ti,ab. 833376

50 (assigned or allocated).ti,ab. 1106339

51 (controlled adj7 (study or design or trial)).ti,ab. 1198366

52 (volunteer or volunteers).ti,ab. 559687

53 human experiment/ 627654

54 trial.ti. 1057532

55 or/36‐54 11489708

56 (random$ adj sampl$ adj7 (cross section$ or questionnaire$1 or survey$ or database$1)).ti,ab. not (comparative study/ or controlled study/ or randomi?ed controlled.ti,ab. or randomly assigned.ti,ab.) 19082

57 Cross‐sectional study/ not (randomized controlled trial/ or controlled clinical study/ or controlled study/ or randomi?ed controlled.ti,ab. or control group$1.ti,ab.) 770008

58 (((case adj control$) and random$) not randomi?ed controlled).ti,ab. 39055

59 (Systematic review not (trial or study)).ti. 442756

60 (nonrandom$ not random$).ti,ab. 36472

61 Random field$.ti,ab. 6101

62 (random cluster adj3 sampl$).ti,ab. 2792

63 (review.ab. and review.pt.) not trial.ti. 2204748

64 we searched.ab. and (review.ti. or review.pt.) 86858

65 update review.ab. 245

66 (databases adj4 searched).ab. 106268

67 (rat or rats or mouse or mice or swine or porcine or murine or sheep or lambs or pigs or piglets or rabbit or rabbits or cat or cats or dog or dogs or cattle or bovine or monkey or monkeys or trout or marmoset$1).ti. and animal experiment/ 1202277

68 Animal experiment/ not (human experiment/ or human/) 2529839

69 or/56‐68 5957727

70 55 not 69 10459196

71 35 and 70 9169

72 11 use cctr 1085

73 22 use ppezv 799

74 71 use emczd 5336

75 72 or 73 or 74 7220

Trial registries

Search date: 8 March 2021 ‐ for setting up the initial JIA

ClinicalTrials.Gov

'juvenile idiopathic arthritis' 269

WHO International Clinical Trials Registry Platform (ICTRP)

'juvenile idiopathic arthritis' 568

Appendix 2. Living reviews

We will republish the review as a new citation version when new evidence has an important impact on the findings of the review; for example, if there is a change in one or more of the following components.

• The findings of one or more major outcomes

• The credibility (e.g. GRADE rating) of one or more major outcomes

• New settings, populations, interventions, comparisons, or outcomes

We will conduct regular search updates and screen for relevant studies.

If we identify no new studies are, the review will remain the same, but the 'What's new' section (or the Updating Classification System, when it is implemented) will be amended to reflect the status, 'No new studies identified with search'.

If the searches return new evidence but the review authors and the editorial team consider it is unlikely to have an important impact on the findings of the review, the 'What's new' section (or the Updating Classification System, when it is implemented) will be amended to reflect the status, 'New information identified, but unlikely to change conclusions'.

The review will be maintained as a living review until the topic is no longer a priority, or we judge that the conclusions of the review will no longer change with the addition of new evidence (Brooker 2019).

Data and analyses

Comparison 1. Methotrexate versus placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Treatment response	3	328	Risk Ratio (M‐H, Random, 95% CI)	1.67 [1.21, 2.31]
1.2 Sustained clinically inactive disease (Wallace criteria or as defined by study authors)	0		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.3 Function (Childhood Health Assessment Questionnaire or equivalent)	0		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.4 Pain (visual analogue scale)	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.5 Participant global assessment of well‐being	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected
1.6 Serious adverse events	3	328	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.04, 8.97]
1.7 Withdrawals due to adverse events	3	328	Risk Ratio (M‐H, Random, 95% CI)	3.46 [0.60, 19.79]

1.2. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 2: Sustained clinically inactive disease (Wallace criteria or as defined by study authors)

1.3. Analysis.

Comparison 1: Methotrexate versus placebo, Outcome 3: Function (Childhood Health Assessment Questionnaire or equivalent)

Comparison 2. Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Treatment response (American College of Rheumatology criteria or as defined by study authors)	0		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
2.2 Sustained clinically inactive disease (Wallace criteria or as defined by study authors)	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
2.3 Function (Childhood Health Assessment Questionnaire or equivalent)	0		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.4 Pain (visual analogue scale or equivalent)	0		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.5 Participant global assessment of well‐being	0		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.6 Serious adverse events	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
2.7 Withdrawals due to adverse events	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

2.1. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 1: Treatment response (American College of Rheumatology criteria or as defined by study authors)

2.3. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 3: Function (Childhood Health Assessment Questionnaire or equivalent)

2.4. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 4: Pain (visual analogue scale or equivalent)

2.5. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 5: Participant global assessment of well‐being

2.6. Analysis.

Comparison 2: Methotrexate (MTX) plus intra‐articular glucocorticoid (IAGC) versus IAGC alone, Outcome 6: Serious adverse events

Comparison 3. Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Treatment response (American College of Rheumatology criteria or as defined by study authors)	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
3.2 Sustained clinically inactive disease (Wallace criteria or as defined by study authors)	0		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
3.3 Function (Childhood Health Assessment Questionnaire or equivalent)	1	94	Mean Difference (IV, Random, 95% CI)	0.05 [‐0.20, 0.30]
3.4 Pain (visual analogue scale or equivalent)	0		Mean Difference (IV, Random, 95% CI)	Subtotals only
3.5 Participant global assessment of well‐being	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
3.6 Serious adverse events	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
3.7 Withdrawals due to adverse events	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

3.2. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 2: Sustained clinically inactive disease (Wallace criteria or as defined by study authors)

3.4. Analysis.

Comparison 3: Methotrexate versus alternative disease‐modifying antirheumatic drug (DMARD), Outcome 4: Pain (visual analogue scale or equivalent)

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bharadwaj 2003.

Study characteristics
Methods	Study design: single‐centre, parallel‐group, 2‐arm, double‐blind randomised placebo‐controlled trial Setting: single site in India Timing: January 1997–June 1998 Interventions: methotrexate 10 mg/m2 versus placebo Sample size calculations: not reported Analysis: Chi2 test for proportion used to compared difference in response rate. Man‐Whitney U test applied for individual disease variables.
Participants	Population 31 participants with JIA (ACR criteria for JRA) enroled 25/31 completed study (14 in methotrexate group, 11 in placebo group) 2/14 dropped out of methotrexate group 4/11 dropped out of placebo group 5/6 dropouts unknown 1/6 dropout due to inadvertent overdose in methotrexate group Inclusion criteria Diagnosis of JRA (ACR criteria) ≥ 4 active joints Unresponsive to 6 weeks of NSAIDs, corticosteroids, or both Exclusion criteria Use of DMARD in preceding 3 months Intercurrent infections Renal derangement Hepatic derangement Baseline characteristics Methotrexate group Age: range 3–18 years Number of participants: 14 Sex: 36% females Disease duration: range 6–84 months JIA subtype proportion: 14% oligoarticular onset; 64% polyarticular; 21.4% systemic Baseline outcome measures not reported Placebo group Age: range 2–22 years Number of participants: 11 Sex: 45% females Disease duration: range 6–108 months JIA subtype proportion: 9% oligoarticular onset; 63.6% polyarticular; 27% systemic onset Baseline outcome measures not reported Pretreatment group differences: groups reasonably well‐matched
Interventions	Experimental group: oral methotrexate 10 mg/m2/week Control group: oral placebo (identical appearance and packaging) Both groups received NSAIDs and (if required) corticosteroids at a dose of 0.5 mg/kg/day Treatment duration: 6 months
Outcomes	Outcomes were measured at baseline, at 3 months, and at 6 months of follow‐up. Primary outcomes measured WHO/ILAR response (graded as nil, 30%, or 50% response) Secondary outcomes measured Pain (VAS) Duration of morning stiffness Number of tender and swollen joints Steinbroker's functional grade Systemic symptoms Global improvement according to parent and physician Haemoglobin ESR Adverse events monitored at each visit Leukocyte count Platelet count Serum transaminase level Urea and creatinine Duration of morning stiffness, pain, patient global assessment, number of swollen joints, number of tender joints, and haemoglobin were reported as median percentage change from baseline. No other prespecified secondary outcomes were reported. Outcomes included in this review at 6 months Treatment response (measured by 30% WHO/ILAR response) Serious adverse events Withdrawals due to adverse event
Notes	Source of funding: not reported Trial registration: not reported Serious adverse events Methotrexate:1 participant (7.1%) inadvertently had an overdose of 24 tablets, requiring supportive management and leucovorin with complete recovery. We judged this would constitute a serious adverse event. Placebo: none reported Withdrawals due to adverse events Methotrexate: 1 participant (7.1%). The participant with the inadvertent overdose withdrew from the study as a consequence. Placebo: none reported Total adverse events Methotrexate: nausea/vomiting in 2 participants, anaemia in 1 participant, elevation of transaminase in 1 participant, and lymphadenopathy in 1 participant. Placebo: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Placebo controlled; placebo looked identical to treatment.
Blinding of outcome assessment (detection bias) Self‐reported outcomes	Low risk	Participant assessments identical between groups.
Blinding of outcome assessment (detection bias)	Low risk	Physician assessments identical between groups and blinded to treatment allocation.
Incomplete outcome data (attrition bias)	Low risk	Missing outcome data balanced in number across intervention groups, with similar reasons for missing data across groups.
Selective reporting (reporting bias)	High risk	No published protocol and unvalidated outcome measure used. Various outcomes measured but not reported.
Other bias	Low risk	None apparent.

Giannini 1992.

Study characteristics
Methods	Study design: multicentre, randomised, double‐blind, placebo‐controlled study Setting: multiple sites in the USA and (the former) Soviet Union Timing: not reported Interventions: methotrexate 5 mg/m2/week versus methotrexate 10 mg/m2/week versus placebo Sample size calculations: not reported Analysis: ITT analysis used; the ITT technique used values of response variables at the final visit, whether or not the participant completed the entire trial
Participants	Population 127 participants aged 18 months–18 years with JIA (ACR criteria for JRA or equivalent criteria used in the Soviet Union and Eastern Europe) were enroled in the study 46 were randomised to low‐dose methotrexate 40 were randomised to very low‐dose methotrexate 41 were randomised to placebo 114 (90%) qualified for analysis of efficacy, with 13 excluded due to violation of study protocol (violation of NSAID dose, prednisone regime, or noncompliance) Inclusion criteria Met ACR criteria for JRA or equivalent criteria used in the Soviet Union & Eastern Europe ≥ 3 joints with active arthritis not adequately controlled by NSAIDS or 2nd‐line agents Age 18 months–18 years Exclusion criteria Presence of clinically important severe or other chronic disease Pregnancy Previous use of methotrexate Baseline characteristics Low‐dose methotrexate (10 mg/m2) group Age: mean 10.1 years Number of participants: 46 Sex: 72% females Disease duration: mean 57.6 months JIA subtype proportion: 9/46 (20%) systemic onset disease Number (%) taking low‐dose prednisone: 15/46 (33%) Number (%) taking 2 NSAIDs: 5/46 (11%) Number of joints: mean 27 (SE 2) Methotrexate very low‐dose (5 mg/m2) group Age: mean 9.6 years Number of participants: 40 Sex: 73% females Disease duration: mean 57.6 months JIA subtype proportion: 11/40 (28%) systemic onset disease Number (%) taking low‐dose prednisone: 15/40 (37%) Number (%) taking 2 NSAIDs: 3/40 (7.5%) Number of joints: mean 21 (SE 2) Placebo group Age: mean 10.6 years Number of participants: 41 Sex: 83% females Disease duration: mean 69.6 months JIA subtype proportion: 12/41 (29%) systemic onset disease Number (%) taking low‐dose prednisone: 14/41 (34%) Number (%) taking 2 NSAIDs: 3/41 (7.3%) Number of joints: mean 24 (SE 2) Pretreatment group differences: no significant differences among the treatment groups in the demographics or disease characteristics
Interventions	Experimental group 1: low‐dose methotrexate (10 mg/m2/week) Experimental group 2: very low‐dose methotrexate (5 mg/m2/week) Control group: oral placebo (identical appearance) Permitted medications had to be constant 1 month prior to trial and not be changed during trial. Maximum of 2 NSAIDs, and prednisolone dose not exceeding 0.5 mg/kg/day (max 10 mg/day). Treatment Duration: 6 months
Outcomes	Outcomes were measured at baseline and monthly for 6 months (7 visits in total). Primary outcomes measured Treatment response (25% composite index) based on articular severity score, which was a sum of the severity scores for swelling, pain on motion, tenderness, and limitation of motion, calculated for each joint and then totalled). Secondary outcomes measured Joint count (swelling) Joint count (tenderness) Joint count (limitation of movement) Joint count (active arthritis) Severity of swelling (none, mild, moderate, severe) Severity of pain on motion (none, mild, moderate, severe) Severity of tenderness (none, mild, moderate, severe) Severity of limitation of motion (full range, 25% limitation, 50% limitation, 75% limitation, no motion possible) Duration of morning stiffness Articular severity score: sum of severity ratings of swelling, pain on motion, tenderness, and limitation of motion Outcomes used in this review Combined score of the low‐dose and very low‐dose methotrexate groups Treatment response using the study's own 25% composite index Pain derived from 'severity of pain on motion' outcome Serious adverse events Withdrawals due to adverse events
Notes	Funding: this study was conducted under the Cooperation in Medical Science and Public Health Agreement and was a collaboration between the USA and the Soviet Union. It was supported by a grant from the Food and Drug Administration, the National Institute of Arthritis and Musculoskeletal and Skin Diseases, by a Clinical Projects Grant from the National Arthritis Foundation, by the Children's Hospital Research Foundation, Cincinnati, and by Lederle Laboratories, Pearl River, NY. Medication was provided by Lederle Laboratories. Trial registration: not reported Adverse events Low‐dose methotrexate Serious adverse events Low‐does methotrexate: 0/46 (0%) Very low‐dose methotrexate: 0/40 (0%) Placebo: 1/41 (2.4%); 1 participant experienced 2 episodes of severe stomach pain graded as a severe adverse event Withdrawals due to adverse events Low‐does methotrexate: 2/46 (4.3%); 1 for LFT derangement, 1 for persistent haematuria. Both resolved quickly after discontinuation of study medication. Very low‐dose methotrexate: 1/40 (2.5%) 1 month after study entry due to persistent skin rash Placebo: 0/41 (0%) Total adverse events Low‐does methotrexate: 6/46 (13%); 3 had gastrointestinal problems alone, 1 had ulceration of mucous membranes, 1 had headache alone, and 1 had headache and abdominal problems. Very low‐dose methotrexate: 1/40 (2.5%) 1 month after study entry due to persistent skin rash Placebo: 5/41 (12%); all side effects were gastrointestinal
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation in blocks of 3 within each centre, with no a priori stratification of participants at entry. All centres followed an identical clinical protocol and used standardised case report forms.
Allocation concealment (selection bias)	Low risk	Off‐site concealment.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Blinding of participants and investigators; methotrexate tablets indistinguishable to placebo tablets; compliance verified by tablet count.
Blinding of outcome assessment (detection bias) Self‐reported outcomes	Low risk	Participants blinded; inclusion criteria appropriate (JRA) for condition being studied.
Blinding of outcome assessment (detection bias)	Low risk	Physicians blinded for physician‐reported outcomes (Physicians' Global Assessment, joint count).
Incomplete outcome data (attrition bias)	Low risk	Emphasis on ITT analysis; ITT analysis used values of response variables at final visit, regardless of whether the participant completed the trial; low dropout rate (108/127 completed the 6‐month trial).
Selective reporting (reporting bias)	Low risk	Mean of numerical data reported (i.e. joint count, articular severity index).
Other bias	Low risk	None apparent.

Ravelli 2017.

Study characteristics
Methods	Study design: prospective, open‐label, randomised, multicentre trial Setting: multiple sites in Italy Timing: July 2009–March 2013 Interventions: IAGC injection(s) and commencement of methotrexate 15 mg/m2/week (max 20 mg/week) orally versus IAGC injection(s) alone (single set of injection(s) at baseline Sample size calculations: based on previous analysis of 94 children with JIA. Assuming withdrawal rate of 5%, the study authors determined that they needed 252 participants to show significance, at 80% power and alpha of 0.05, if methotrexate increased remission rates by 30%. Analysis: all primary and secondary outcomes were assessed by ITT, with analyses including all randomised participants who started treatment.
Participants	Population 226 children/young people with JIA (ILAR criteria) enroled 207/226 randomised; 19 did not consent 105 assigned to IAGC plus methotrexate group; 92 completed study 102 assigned to IAGC alone group; 100 completed study 48/226 (23%) injected in 1 joint and 159/226 (77%) injected in ≥ 2 joints All randomised participants included in ITT analysis Inclusion criteria Age < 18 years ILAR diagnosis of oligoarticular JIA Candidate to receive intra‐articular steroids in ≥ 2 joints or single joint if flare within 12 months after a previous intra‐articular steroid injection Exclusion criteria Knee monoarthritis treated with steroid injection Previous treatment with methotrexate or other synthetic or biologic DMARD Intra‐articular steroids within 3 months before enrolment Baseline characteristics Methotrexate plus IAGC group Age: mean 4.1 years; range 2.4–8.9 years Number of participants: 105 Sex: 80% females Disease duration: mean 6.9 months; range 2.8–19.5 months JIA subtype proportion: oligoarticular JIA in 100% Positive ANA: 73/86 (85%) Previous use of prednisone: 4/105 (4%) Previous IAGC injection: 46/105 (44%) Concurrent use of NSAIDs: 12/105 (11%) Physician's global assessment: mean 5.0 cm range 4.0–6.0 cm ESR: mean 32 mm/h; range 15–54 mm/h CRP: mean 0.5 mg/dL; range 0.5–1.4 mg/dL Number of injected joints: mean 2; range 2–3 Participants with 1 injected joint: 26/105 (25%) Participants with more than 1 injected joint: 79/105 (75%) IAGC alone group Age: mean 2.8 years; range 1.6–6.0 years Number of participants: 102 Sex: 72% females Disease duration: mean 6.7 months; range 2.8–19.5 months JIA subtype proportion: oligoarticular JIA in 100% Positive ANA: 73/87 (84%) Previous use of prednisone: 2/102 (2%) Previous IAGC injection: 45/102 (44%) Concurrent use of NSAIDs: 9/102 (9%) Physician's global assessment: mean 5.0 cm range 4.0–6.3 cm ESR: mean 22 mm/h; range 9–40 mm/h CRP: mean 0.5 mg/dL; range 0.5–1.5 mg/dL Number of injected joints: mean 2; range 2–3 Participants with 1 injected joint: 22/102 (22%) Participants with more than 1 injected joint: 80/102 (78%) Pretreatment group differences: no meaningful differences between treatment groups
Interventions	Experimental group: IAGC injection(s) plus oral methotrexate 15 mg/m2/week (maximum dose 20 mg) Control group: IAGC injection(s) alone Single set of concomitant IAGC injection/s performed at baseline: triamcinolone hexacetonide 1 mg/kg (maximum 40 mg) in knees and shoulders, 0.75 mg/kg (maximum 30 mg) in ankles and elbows, 0.25–0.5 mg/kg (maximum 20 mg) in wrists; methylprednisolone acetate 5–10 mg in small hand and foot joints, 20–40 mg in subtalar and intertarsal joints Participants had to stop using NSAIDs before enrolment. Treatment duration: 12 months
Outcomes	Outcomes were measured at baseline and at 12 months. Primary outcomes measured Proportion of participants in remission (defined as absence of clinical sign of active arthritis in all injected joints) at 12 months Secondary outcomes measured Proportion of participants with relapse of synovitis in specified joints Time to flare Proportion of participants who achieved inactive disease as per Wallace criteria (Wallace 2011) cJADAS (Consolaro 2014) Frequency of subsequent IAGC procedure with or without general anaesthesia Safety Outcomes used in this review at 12 months Sustained clinically inactive disease as per Wallace criteria (Wallace 2011) Serious adverse events Withdrawals due to adverse events
Notes	Funding: funding by the Italian Agency of Drug Evaluation. The funder had no role in study design, data collection, data analysis, data interpretation, or report writing, but did review the final report before submission. Trial registration: European Union Clinical Trials Register (EudraCT) number 2008‐006741‐70 Serious adverse events IAGC + methotrexate: 0/105 (0%) IAGC alone: 0/102 (0%) Withdrawals due to adverse events IAGC + methotrexate: 2/105 (1.9%). 1 participant permanently discontinued treatment due to increased liver transaminases and 1 due to gastrointestinal discomfort. IAGC alone: 0/102 (0%) Total adverse events IAGC + methotrexate: 20/105 (17%). 14 participants experienced gastrointestinal discomfort (nausea, vomiting, or constipation), 8 had increased liver transaminases, 2 experienced fatigue, 2 experienced irritability, 1 experienced hair loss, and 1 had leukopenia. IAGC alone: 0/102 (0%). No adverse effects related to systemic absorption of IAGC were recorded in either group. No participants developed skin hypopigmentation, subcutaneous atrophy, or other localised adverse events at the site of intra‐articular injection in either group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation in 1:1 ratio using computer program (SPSS version 19).
Allocation concealment (selection bias)	Low risk	Off‐site concealment.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and physicians not blinded to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes	High risk	No placebo control, and patient‐reported outcome measures used.
Blinding of outcome assessment (detection bias)	High risk	Blinded assessment not feasible in some participating centres because only 1 paediatric rheumatologist was available.
Incomplete outcome data (attrition bias)	Low risk	Low numbers of participants not included in final analysis (2/102 placebo, 13/105 methotrexate), so unlikely that imputed data affected outcome (best‐ and worst‐case scenarios unlikely to change final outcome).
Selective reporting (reporting bias)	Unclear risk	Unclear if the trial measured adverse events other than those related to steroids in the control group.
Other bias	Low risk	None apparent.

Silverman 2005.

Study characteristics
Methods	Study design: multicentre, double‐blind RCT Setting: multiple sites across 12 countries Timing: March 2002–January 2003 Interventions: oral methotrexate versus oral leflunomide Sample size calculations: using a superiority design, the investigators estimated that 37 patients per group would give a statistical power of 80% to detect an absolute difference between groups of 15% in the Percent Improvement Index at week 16, assuming an SD of 23% and completion rates of 80% (alpha = 0.05). Analysis: ITT
Participants	Population 103 participants were screened 94 underwent randomisation 47 were assigned to methotrexate group 47 were assigned to leflunomide group 44 in methotrexate group completed randomised arm of study 42 in leflunomide completed randomised arm of study Inclusion criteria Age 3–17 years ACR criteria for JRA Polyarticular course disease No previous treatment with methotrexate or leflunomide Exclusion criteria ACR functional class IV disease Active systemic symptoms within 4 weeks before entry Persistent or severe infection within 3 months before entry Current inflammatory disease other than JRA Baseline characteristics Methotrexate group Age: mean 10.2 years Number of participants: 47 Sex: 72% females Disease duration: mean 16.44 months JIA subtype proportion: 39/47 (83%) polyarticular JIA; 8/47 (17%) pauciarticular JIA; 0/47 (0%) systemic onset ANA positivity: 19/47 (40%) Rheumatoid factor‐positive: 9/47 (19%) Number of active joints: mean 14.0 (SD 9.9) Number of joints with limited range of motion: mean 8.0 (SD 6.6) Physician global assessment (0–100 mm): mean 47.3 (SD 19.3) Patient global assessment (0–100 mm): mean 36.5 (SD 23.8) CHAQ (0–3): mean 1.11 (SD 0.74) ESR: mean 34.5 mm/hr (SD 21.7) CRP: mean 13.8 mg/L (SD 25.6) No history of DMARD therapy: 43/47 (91%) Corticosteroid use at study entry: 9/47 (19%) Leflunomide group Age: mean 10.1 years Number of participants: 47 Sex: 75% females Disease duration: mean 20.28 months JIA subtype proportion: 77% polyarticular JIA; 21% pauciarticular JIA; 1 (2%) systemic onset ANA positivity: 13/47 (28%) Rheumatoid factor‐positive: 10/47 (21%) Number of active joints: mean 14.4 (SD 7.9) Number of joints with limited range of motion: mean 7.7 (SD 6.4) Physician global assessment (0–100 mm): mean 55.1 (SD 18.3) Patient global assessment (0–100 mm): mean 39.6 (SD 28.1) CHAQ (0–3): mean 1.03 (SD 0.71) ESR: mean 30.8 mm/hr (SD 18.2) CRP: mean 19.6 mg/L (SD 22.8) No history of DMARD therapy: 45/47 (96%) Corticosteroid use at study entry: 9/47 (19%) Pretreatment group differences: no meaningful difference between treatment groups
Interventions	Experimental group: oral methotrexate 0.5 mg/kg/week plus oral matching placebo Control group: oral leflunomide with a dose dependant on bodyweight (20 kg: 100 mg for 1 day, then 10 mg every other day maintenance dose; 20~40 kg: 100 mg for 2 days followed by 10 mg/day; > 40 kg: 100 mg for 3 days followed by 20 mg/day maintenance dose) All participants received at least 5 mg folate per week. Up to 2 intraarticular steroid injections (triamcinolone hexacetonide) was allowed per participant during the study. Participants could have concomitant treatment with NSAIDs and no more than 0.2 mg of prednisone or the equivalent per kg of bodyweight per day (maximum daily dose 10 mg), provided the dose was unchanged at least 2 weeks before enrolment and throughout the study. Treatment duration: 16‐week randomised arm, with subsequent offer of blinded 32‐week treatment extension arm
Outcomes	Primary outcomes measured Mean percentage improvement index Percentage of participants with a 30% improvement from baseline in ≥3/6 core variables (pedACR30). The core set variables are: physician global assessment of disease activity (scored on a 10‐cm VAS); parent/patient global assessment of overall well‐being (scored on a 10‐cm VAS); functional ability (CHAQ); number of joints with active arthritis; number of joints with limited range of motion; and ESR. Secondary outcomes measured pedACR50: 50% improvement in ≥3/6 core variables pedACR70: 70% improvement in ≥3/6 core variables Time to pedACR30 response Mean change in each of the core set variables described above CRP Outcomes used in this review at 16 weeks Treatment response, based on pedACR70 Function (CHAQ) Patient/parent global assessment of overall well‐being (0–10 cm VAS) Serious adverse events Withdrawals due to adverse events
Notes	Source of funding: supported by funding by Sanofi‐Aventis Trial registration: not reported Serious adverse events Methotrexate group: 0/47 (0%) Leflunomide group: 3/47 (6.4%) Withdrawals due to adverse events Methotrexate group: 1/47 (2.1%). 1 participant withdrew due to liver enzyme dysfunction. Leflunomide group: 3/47 (6.4%). 1 participant withdrew due to liver dysfunction, 1 participant due to development of parapsoriasis, and 1 participant withdrew due to development of Crohn's disease. Total adverse events Methotrexate group: 38/47 (81%). A wide range of adverse events were reported, the most common of which were headache (23%), nasopharyngeal symptoms (23%), nausea (34%), diarrhoea (17%), and upper respiratory tract infection (13%). Leflunomide group: 43/47 (91%). A wide range of adverse events were reported, the most common of which were headache (38%), nasopharyngeal symptoms (32%), abdominal pain (26%), nausea (28%), diarrhoea (15%), and alopecia (15%).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Centralised randomisation; similar baseline data.
Allocation concealment (selection bias)	Low risk	Off‐site concealment.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participant and assessor blinded.
Blinding of outcome assessment (detection bias) Self‐reported outcomes	Low risk	Participant unaware of treatment allocation.
Blinding of outcome assessment (detection bias)	Low risk	Assessor unaware of treatment allocation.
Incomplete outcome data (attrition bias)	Low risk	Low discontinuation rate between group analysis; the last observation was carried forward in the case of missing data.
Selective reporting (reporting bias)	Low risk	Study protocol unavailable, but commonly used core‐set outcomes are published.
Other bias	Low risk	None apparent.

Woo 2000.

Study characteristics
Methods	Study design: multicentre, double‐blind, randomised controlled cross‐over trial Setting: multiple sites in the UK and France Timing: not reported Interventions: oral methotrexate versus oral placebo Sample size calculations: not reported Analysis: ITT
Participants	Population 88 people were enrolled. This was a cross‐over study where all participants alternated through both a placebo and treatment phase, with randomisation occurring to which phase of treatment a participant is assigned to. All 88 participants, including withdrawals, were analysed Inclusion criteria Age <16 years ILAR criteria for extended oligoarthritis or systemic arthritis For children with systemic arthritis: minimum of 3 months' disease duration for children with systemic arthritis For children with systemic arthritis: oral steroids in doses > 5 mg in addition to NSAID therapy For children with extended oligoarthritis: active polyarthritis not responding to NSAID or local steroid treatment for at least 1 year Exclusion criteria Abnormal results on liver, kidneys, or chest radiograph Diabetes Infection Thrombocytopenia (< 50,000) Leukopenia (< 30,000) Prior treatment with methotrexate or other cytotoxic drugs Baseline characteristics The study authors described baseline characteristics of separate JIA subtypes, rather than the baseline characteristics between the intervention and placebo groups. Pretreatment group differences Because of the cross‐over design of the study, there would be no meaningful difference between the groups studied.
Interventions	Experimental group: oral methotrexate 15 mg/m2/week Control group: placebo with identical appearance supplied by Lederle Laboratories. Treatment Duration: 4‐month active/placebo treatment period, followed by 2‐month washout, then 4‐month placebo/active treatment period, followed by another 2‐month washout
Outcomes	Outcomes were measured based on a treatment schedule consisting of an initial 4‐month active/placebo treatment period, followed by a 2‐month washout period, and then a subsequent 4‐month placebo/active treatment period. Thus, outcomes were measured at 4 months and 10 months. Primary outcomes measured Physician global assessment of disease activity (0–10‐cm VAS) Participant global assessment (0–10‐cm VAS) Number of joints with active disease Range of joint motion CRP ESR Overall clinical improvement reported as a > 30% response using Giannini's criteria (Giannini 1997) Secondary outcomes measured Haemoglobin Platelet count White cell count Total protein IgG levels. Outcomes used in this review at 4 months (before cross‐over) Treatment response (30% response of 3/5 core set variables using Giannini's criteria) Parent global assessment of well‐being (0–10 cm VAS) Serious adverse events Withdrawals due to adverse event
Notes	Funding: medication was provided by Lederle Laboratories. Trial registration: not reported Serious adverse events Methotrexate: 0/88 (0%) Placebo: 0/88 (0%) Withdrawals due to adverse events Methotrexate: 2/88 (2.2%). 1 because of an increase in liver enzyme levels during the MTX treatment period, 1 due to an intercurrent hepatitis A infection. Placebo: 0/88 (0%) Total adverse events: the study did not specify the total number of adverse events, though there were reports of adverse events of various types occurring in both methotrexate and placebo cross‐over periods. They included nausea, gastrointestinal upset, mouth ulcers, hair loss, mood change, pneumonitis, bone marrow failure, and abnormal liver function tests.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation done in 2 blocks in 2 centres with treatment dispensed by the pharmacist at each centre.
Allocation concealment (selection bias)	Low risk	Off‐site concealment.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Blinding of participants and key study personnel ensured, and unlikely that blinding could have been broken.
Blinding of outcome assessment (detection bias) Self‐reported outcomes	Low risk	Participants unaware of treatment.
Blinding of outcome assessment (detection bias)	Low risk	Physicians blinded for global assessment and joint count assessments.
Incomplete outcome data (attrition bias)	Low risk	Similar proportions lost between groups.
Selective reporting (reporting bias)	Unclear risk	No published protocol, no core outcome sets available at time of publication, although published results fit with prespecified aims.
Other bias	Low risk	None apparent.

ACR American College of Rheumatology; ANA antinuclear antibody; CHAQ Childhood Health Assessment Questionnaire; cJADAS clinical Juvenile Arthritis Disease Activity Score; CRP C‐reactive protein; DMARD disease‐modifying antirheumatic drug; ESR erythrocyte sedimentation rate; EudraCT: European Union Drug Regulating Authorities Clinical Trials Database; IAGC: intra‐articular glucocorticoid; IgG: immunoglobulin G; ILAR: International League Against Rheumatism; ITT: intention‐to‐treat; JIA: Juvenile idiopathic arthritis; JRA juvenile rheumatoid arthritis; NSAID nonsteroidal anti‐inflammatory drugs; NY: New York; pedACR70/50/30: American College of Rheumatology paediatric 70, 30, 50 criteria; RCT: randomised controlled trial; SE: standard error; SPSS: Statistical Package for the Social Sciences; VAS: visual analogue scale; WHO: World Health Organization.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Dupuis 1995	Ineligible intervention (methotrexate fasted versus methotrexate not fasted).
EUDRACT 2018‐001571‐21	Ineligible intervention (psychological intervention).
Foell 2010	Ineligible intervention (methotrexate withdrawal at 6 months versus methotrexate withdrawal at 12 months).
Forster 2000	Ineligible intervention (intra‐articular glucocorticoids versus placebo).
Gao 2003	Ineligible intervention (methotrexate versus methotrexate plus leflunomide).
Hissink 2019	Ineligible intervention (methotrexate or sulfasalazine versus methotrexate/glucocorticoids versus methotrexate/etanercept).
Kvien 1985	Ineligible intervention (hydroxychloroquine versus gold sodium thiomalate versus penicillamine).
NCT04614311	Ineligible intervention (intra‐articular glucocorticoid versus no intraarticular glucocorticoid).
NL5742	Ineligible intervention (methotrexate versus methotrexate/sulfasalazine/hydroxychloroquine).
Rezaieyazdi 2023	Ineligible intervention (methotrexate and leflunomide versus methotrexate and placebo).
Romero 2017	Ineligible study population (adults).
Ruperto 2004	Ineligible intervention (methotrexate intermediate dose versus methotrexate high dose).
Tynjälä 2011	Ineligible intervention (methotrexate versus methotrexate/sulfasalazine/hydroxychloroquine versus methotrexate/infliximab).
Verkaaik 2011	Ineligible intervention (methotrexate oral versus methotrexate subcutaneously versus methotrexate oral/behavioural therapy).
Wallace 2012	Ineligible intervention (methotrexate plus placebo versus methotrexate/etanercept/prednisone).

Characteristics of studies awaiting classification [ordered by study ID]

Burrone 2022.

Methods	Study design: randomised, parallel assignment, interventional Setting: "... all Italian centres belonging to the Paediatric Rheumatology International Trials Organization (PRINTO) were invited to participate in the study. In January 2022, after a feasibility survey, PRINTO centers outside Italy were also invited to participate ..." Timing: recruitment commenced 21 May 2019 Interventions: 'Step up' treatment protocol versus 'Step down' treatment protocol Sample size calculations: extrapolated from previous study on early aggressive therapy on JIA patients (Wallace 2012) Analysis: ITT
Participants	Planned sample size: estimated enrolment of 260 participants Inclusion criteria Recent‐onset JIA with oligoarthritis or RF‐negative polyarthritis (ILAR classification) Age 2–17 years DMARD‐naive (only treatment with 1 NSAID is allowed for no more than 6 weeks from diagnosis) No corticosteroid joint injections in the 3 months before enrolment Exclusion criteria Classification in 1 of the following JIA categories: systemic arthritis, RF‐positive polyarthritis, psoriatic arthritis, enthesitis‐related arthritis, undifferentiated arthritis Need for systemic treatment for uveitis Tuberculosis‐related issues Previous treatment with any synthetic or biologic DMARD Any live attenuated vaccine within 4 weeks prior to the baseline visit Prior or current history of malignancy or any other significant concomitant illness Specified laboratory derangement on complete blood count or liver enzymes
Interventions	Step‐up treatment group: conventional strategy based on treatment escalation and driven by the treat‐to‐target strategy. Less severe participants will receive IAGC injections without systemic therapy. If the JADAS10 state of minimal disease activity (MiDA) after 3 months or the JADAS10 inactive disease state at 6 months are not reached, treatment with methotrexate will be started, preferably subcutaneously in a single weekly dose of 15 mg/m2 (max 20 mg), and IAGC injections can be repeated. Then, if the states of JADAS MiDA after 3 months or JADAS inactive disease at 6 months are not reached, treatment with an anti‐TNF agent in a single weekly dose of 0.8 mg/kg (max 50 mg) subcutaneously will be commenced. More severe participants in the Step‐up arm will receive IAGCs and MTX as a first step and then escalate therapy with the same pathway, adding an anti‐TNF treatment as a second step and switching to a different biologic medication as a third step. Step‐down treatment group: combined, aggressive therapy involving early combined treatment (methotrexate plus IAGC for less severe oligoarthritis, methotrexate plus etanercept for severe oligoarthritis and polyarthritis). After 6 months, if JADAS10 ID is achieved, less severe patients will have methotrexate discontinued, whereas more severe patients will stop etanercept and continue methotrexate. Children in the Step‐down arm experiencing a worsening from the baseline of the JADAS10 at 3 months, not achieving JADAS10 inactive disease at month 6, or losing the state of inactive disease after month 6 will be considered treatment failures and will change treatment according to the attending physician decision. Treatment duration The effectiveness of the 2 treatment strategies will be compared at 12 months.
Outcomes	Outcomes will be assessed at 3, 6, 9, and 12 months. Primary outcomes Clinical remission on or off medication at 12 months Secondary outcomes Inactive disease Time to inactive disease as per JADAS/JIA ACR criteria Time to JADAS/JIA ACR clinical remission Time spent on JADAS/JIA ACR inactive disease Cumulative level of disease activity throughout the study period Time spent on therapy Rate of flares Rate of uveitis onset
Notes	Trial registration: "The Trial is registered on the ClinicalTrials.gov registry (NCT03728478) on the 31st October 2018 and EU Clinical Trials Register on the 14th May 2018 (Eudract Number: 2018–001931‐27)." Funding: "The study is funded by The Italian Agency of Drug Evaluation (Contract number AIFA‐2016‐02364494), Compagnia di San Paolo (ID: 2017.0657) and IRCCS Istituto Giannina Gaslini, Genoa, Italy. Pfizer Srl provided etanercept that is distributed to the participating sites in Italy, for Step‐down arm patients (Investigator Initiated Research Grant WI236538)." Declarations of interest: "Marco Burrone, Marta Mazzoni, Roberta Naddei, Angela Pistorio, Maddalena Spelta, Marco Garrone, Maria Lombardi, Elisa Patrone, Silvia Scala, Luca Villa have no conflicts of interest to disclose. Nicolino Ruperto has received honoraria for consultancies or speaker bureaus from the following pharmaceutical companies in the past 3 years: 2 Bridge, Amgen, AstraZeneca, Aurinia, Bayer, Brystol Myers and Squibb, Celgene, inMed, Cambridge Healthcare Research, Domain Therapeutic, EMD Serono, Glaxo Smith Kline, Idorsia, Janssen, Eli Lilly, Novartis, Pfizer, Sobi, UCB. The IRCCS Istituto Giannina Gaslini (IGG), where NR works as full‐time public employee, has received contributions from the following industries in the last 3 years: Bristol Myers and Squibb, Eli‐Lilly, F Hofmann‐La Roche, Novartis, Pfizer, Sobi. This funding has been reinvested for the research activities of the hospital in a fully independent manner, without any commitment with third parties. Angelo Ravelli has received honoraria for consultancies or speaker bureaus from the following pharmaceutical companies in the past 3 years: AbbVie, Angelini, BMS, Pfizer, Hofman LaRoche, Novartis, Pfizer, Reckitt Benckiser. Alessandro Consolaro reports Investigator initiated research grant from Pfizer and Alfa Sigma and speaker's bureaus from Pfizer."

ACR: American College of Rheumatology; DMARD disease‐modifying antirheumatic drug; ILAR: International League of Associations for Rheumatology; ITT: intention‐to‐treat; JADAS Juvenile Arthritis Disease Activity Score; IAGC: intra‐articular glucocorticoid; JIA: juvenile idiopathic arthritis; MTX: methotrexate; NSAID nonsteroidal anti‐inflammatory drug; RF: rheumatoid factor; TNF: tumour necrosis factor.

Differences between protocol and review

Since the last update of this review (Takken 2001), we have updated the methods to conform with updated conduct and reporting standards of systematic reviews, as recommended by Cochrane's MECIR project and in accordance with current Cochrane Musculoskeletal guidance.

We added a comparison: methotrexate in addition to intra‐articular glucocorticoids compared with intra‐articular glucocorticoids alone. We had not anticipated this comparison, but we identified a study that the expert Panel for the corresponding living guideline advised was relevant for this review.

We had initially planned to use Cochrane Crowd to tag studies as RCTs, but we did not use it because the initial search yield was manageable.

We had initially planned to extract outcome measures that assessed benefits of treatment at multiple time points: up to six months, up to 12 months (primary time point for the primary comparison), and over 12 months. And we had planned to extract data on withdrawals due to adverse events and serious adverse events at the end of the trial. However, as only one trial reported outcomes at 12 months, and the remaining four reported their final measurements at four or six months, we altered the primary time point to be the final time point as reported in the trial.

In the 'Plain language summary' column of the summary of findings table, we had initially planned to report the absolute percent difference and the relative percent change from baseline, but the Review Group recommended we include a statement interpreting the results.

We updated outcomes to conform with contemporary OMERACT‐endorsed core domains.

Contributions of authors

JT: co‐ordination of the review; search and selection of studies for inclusion in the review; collection of data for the review; assessment of the risk of bias in the included studies; analysis of data; assessment of the certainty in the body of evidence; interpretation of data; and writing of the review.
WR: design of the updated review; search and selection of studies for inclusion in the review; collection of data for the review; assessment of the risk of bias in the included studies; analysis of data; assessment of the certainty in the body of evidence; interpretation of data; and writing of the review.
SW: design of the updated review; assessment of the certainty in the body of evidence; interpretation of data; and writing of the review.
TT: conception of the review; design of the original review; writing of the review.
RJ: design of the updated review; co‐ordination of the review; arbitration during study selection, data extraction, and assessment of the risk of bias in the included studies; analysis of data; assessment of the certainty in the body of evidence; interpretation of data; and writing of the review.
GT: interpretation of data; and writing of the review.
JM: interpretation of data; and writing of the review.
RB: design of the updated review; interpretation of data; writing of the review; and acting as guarantor for the review.

Sources of support

Internal sources

• Department of Musculoskeletal Health, School of Public Health and Preventive Medicine, Monash University, Australia

  In kind support.

External sources

• Australia and New Zealand Musculoskeletal (ANZMUSC) Clinical Trial Network NHMRC Centre of Research Excellence, Australia

  SL Whittle is supported, in part, by an Australia and New Zealand Musculoskeletal (ANZMUSC) Clinical Trial Network Practitioner Fellowship

• National Health and Medical Research Council, Australia

  R Buchbinder is supported by an Australian National Health and Medical Research Council Investigator Fellowship.

Declarations of interest

JT works as a Paediatric Rheumatologist; the intervention in question is widely used in his routine clinical practice. He declares personal payments from Pfizer Australia*: consultancy fees for participating in a Steering Committee for a rheumatology education meeting (EXcite) from 8 to 10 of September 2023, as well as travel and accommodation expenses for attending the meeting.

WR works as a Paediatric Rheumatologist at The Royal Children's Hospital in Melbourne and The Monash Children's Hospital in Melbourne. WR declares that he is a principal investigator for the ADJUST trial (NCT03816397), which is an ongoing study that is funded by the National Institute of Health (NIH); the investigational medicinal product is supplied by AbbVie. This is an unpaid position; however, WR has received support for travel and accommodation to attend an investigator meeting in San Francisco in 2019; paid to institution. WR also declares that he has published opinions on the topic (DOI 10.1186/s12969-019-0366-x; DOI 10.1093/rheumatology/kez441; and DOI 10.1136/archdischild-2018-315819).

SW is a Senior Consultant Rheumatologist at The Queen Elizabeth Hospital and the Vice President of the Australian Rheumatology Association, which has declared an opinion or position on the topic. SW is also a Cochrane Musculoskeletal Editor. He was not involved in the editorial process for this review. He is supported, in part, by an Australia and New Zealand Musculoskeletal (ANZMUSC) Clinical Trial Network Practitioner Fellowship.

TT declares that he has no conflicts of interest.

RJ is the Managing Editor for Cochrane Musculoskeletal. RJ was involved in selecting peer reviewers but was not otherwise involved in the editorial process for this review.

GT declares income from private practice.

JM declares income from private practice: she is the business owner of a private practice, the Victorian Children's Clinic, since 2012. JM also declares personal payments from AbbVie, for a one‐hour education panel on clinician well‐being in 2020 (one‐off fee), and from Pfizer Australia*, for giving a talk at the Excite meeting in September 2023.

Rachelle Buchbinder works as a Rheumatologist (adult) and is the Co‐ordinating Editor for Cochrane Musculoskeletal. She was not involved in the editorial process for this review. She is supported by an Australian National Health and Medical Research Council Investigator Fellowship.

*Review authors acquired these conflicts of interest after completion and submission of the review to the Central Editorial Service. Given the time points involved, the Research Integrity Team decided that these interests were not in breach of the Cochrane Conflict of Interest policy.

New search for studies and content updated (no change to conclusions)