2011 American College of Rheumatology recommendations for the treatment of juvenile idiopathic arthritis: Initiation and safety monitoring of therapeutic agents for the treatment of arthritis and systemic features^{†‡§¶‖†}

RAND/UCLA Appropriateness Method overview.

The RAND/UCLA Appropriateness Method was originally developed to help determine when the benefits of a medical intervention outweigh the risks (18) with the understanding that the published literature often does not provide evidence at the level of detail required to guide decisions in everyday clinical practice. This method relies upon the efforts of two distinct groups of participants: the Core Expert Panel (CEP) and the Task Force Panel (TFP). Our CEP was composed of experienced pediatric rheumatologists from the US, Canada, and Europe who are among the world's leaders in the investigation of the treatment of JIA. The TFP contained internationally recognized pediatric rheumatology clinicians and researchers from the US, Canada, and Europe; an advanced practice pediatric rheumatology nurse; a general pediatrician with expertise in evidence-based medicine; and a patient representative, i.e., a parent of a child with JIA who has broad experience in JIA and family treatment preferences.

The initial step was a systematic review of the literature, which was then used by the CEP to prepare a summary report of the latest scientific evidence. Concurrently, the CEP prepared a comprehensive list of clinical scenarios (or potential indications) for each medical intervention of interest to be addressed by the recommendations. The scenarios categorized hypothetical patients using all possible combinations of key clinical parameters, such as disease activity and prognostic features, relevant to the decision process. The evidence report and the clinical scenarios were presented to the TFP for their review.

Evaluation of the scenarios by the TFP led directly to the recommendations for each medical intervention. In the first round of voting, the TFP anonymously and independently rated the appropriateness of the medical interventions in the clinical scenarios based on the scientific evidence and their best clinical judgment. Differences in opinion from the first round of voting were discussed at a face-to-face meeting, followed by a second rating of the clinical scenarios. The final ratings were compiled and assessed by the CEP and no significant discrepancies were identified. Finally, the ratings were used by the CEP to create recommendations from the clinical scenarios.

Scope of recommendations.

These recommendations cover the indications and safety monitoring for the use of NSAIDs, intraarticular glucocorticoid injections, nonbiologic DMARDs, biologic DMARDs, and systemic glucocorticoids for the treatment of the systemic features of systemic arthritis. In contrast to prior ACR recommendations, the scope of this project was not explicitly established by the ACR, but rather by the CEP members. An exhaustive list of broad potential topics for inclusion was constructed by the principal investigator (TB) and reviewed and revised by 3 other authors (RQC, EMD, KGS). The resultant list of 23 potential topics for consideration was distributed to all CEP members. Using a modified Delphi process via electronic mail, the list of topics was prioritized and gradually shortened until it was deemed feasible in the time allotted to the project. As a result, several relevant topics were necessarily omitted. Medication contraindications and intolerance were not considered. We suggest referring to the 2008 ACR recommendations for the use of DMARDs in rheumatoid arthritis (19) for general guidance about medication contraindications. Tapering or discontinuation of medications for patients with inactive disease was also not considered. Only the most relevant and frequently used agents within each medication class were included, as determined using a structured iterative process via electronic mail with repeated revisions to the proposed list until it was accepted by all CEP members. Agents not widely commercially available for the treatment of JIA at the time of the literature search (e.g., canakinumab, rilonacept, tocilizumab) were excluded. The indications for systemic glucocorticoids for the treatment of synovitis were not considered, owing to a lack of published evidence. The treatment of uveitis, enthesitis, and macrophage activation syndrome was not considered.

Systematic literature review.

The systematic literature review was restricted to publications available in Medline. Briefly, we searched using PubMed for articles in English, with abstracts, and published from 1966 to the present. The final search strategy, developed in consultation with a virtual reference librarian, is shown in Supplementary Appendix A (available in the online version of this article at http://onlinelibrary.wiley.com.bibliotheek.ehb.be/journal/10.1002/(ISSN)2151-4658). We performed the final search on March 3, 2009, and identified 756 articles for further consideration.

We next performed title and abstract review of the identified articles. In keeping with the scope of the recommendations, all studies that did not explicitly address any pharmacotherapeutic clinical outcomes of the medications of interest were excluded (e.g., radiographic descriptive studies, pharmacokinetics, genomics, etc.). Because of the relatively small number of identified articles, we included uncontrolled studies, regardless of the number of patients reported. Review articles were excluded. Two reviewers reviewed all of the abstracts for inclusion, blinded to each other's determinations. Disagreement was resolved by a third reviewer (ST-R) and confirmed by a fourth reviewer (TB), when necessary. Following title and abstract review, 239 articles remained for consideration.

Each of the full-length articles was abstracted by one of a team of reviewers with key elements from each article entered into an electronic database. Articles were excluded during the full-length article review when no pharmacotherapeutic clinical outcomes of medications of interest were identified; all of these exclusions were verified by a second reviewer (TB). As a result, 214 articles were included in the final comprehensive evidence report that was presented to the TFP prior to the first round of voting.

An identical literature search was repeated on October 5, 2009, to identify any new publications since the first systematic review. Thirty new articles were identified and subjected to the same review process as the initial search. Seven fully abstracted articles were added to the updated evidence report and presented to the TFP prior to the face-to-face meeting.

Scenario definitions.

In order to develop clinical scenarios in accordance with the RAND/UCLA Appropriateness Method (18), several key clinical decision parameters were deemed necessary by the CEP: disease phenotype, prognostic features, disease activity, and current therapy. Definitions and values for these key parameters were as evidence based as possible and were determined using a structured iterative process via electronic mail with repeated revisions to the proposed parameters until they were accepted by all CEP members.

JIA treatment groups.

Chronic childhood arthritis is a heterogeneous condition. The most recent ILAR disease classification criteria (1) divide JIA into 6 distinct categories. However, this disease classification system was not strictly applied to the development of these recommendations for two reasons: currently, there is minimal evidence to support the differential treatment of children with JIA for many of the category distinctions, and the inclusion of all 6 categories would unnecessarily increase the total number of scenarios to an unmanageable number for consideration by the TFP (18).

In place of the ILAR JIA classification, the CEP developed “treatment groups” for these recommendations with the goal of succinctly representing clinical decision making in the treatment of JIA. The JIA category of systemic arthritis proved to be especially challenging to evaluate. Attempts to exhaustively depict the myriad possible clinical presentations of systemic arthritis were impractical. Therefore, recommendations for the treatment of significant active systemic features (e.g., fever) and active arthritis for patients with systemic arthritis were considered separately and independently by the TFP. Other authors have suggested this dichotomy when considering therapeutic choices for systemic arthritis (22). The appropriate treatment of systemic arthritis patients with concurrently active systemic features and active arthritis may be expected to incorporate elements of both sets of recommendations, but this was not explicitly considered by the TFP. The 5 JIA treatment groups used in these recommendations are described below.

Features of poor prognosis.

Risk stratification is crucial for guiding optimal treatment. The prognosis for patients with JIA is variable, and some clinical factors have been shown to predict worse outcomes. The CEP considered the published literature and their personal clinical experience in developing the features of poor prognosis for these recommendations.

Each JIA treatment group has its own respective features of poor prognosis. In each case, the presence of one listed feature is sufficient to classify the patient as having a poor prognosis for the purposes of these recommendations. A more complex approach to features of poor prognosis would have increased the number of scenarios to an unmanageable number for consideration by the TFP (18). The features of poor prognosis (and their corresponding literature references) are listed according to each JIA treatment group in Tables 1–5. Of note, the CEP believed there was insufficient published evidence available to include damage detected by imaging other than radiographs as a feature of poor prognosis in JIA.

JIA disease activity.

Recommendations require clear definitions of disease activity to make rational therapeutic choices. In these recommendations, “active joints” and “active arthritis” are defined by joints with swelling not due to deformity or joints with limitation of motion and with pain or tenderness (10). “Active fever” means current fever that is attributable to systemic arthritis disease activity. At least one continuous measure of JIA disease activity has been recently developed (39). Nevertheless, currently no continuous disease activity measures have been extensively validated or widely applied to daily clinical practice. Accordingly, the CEP developed levels of disease activity for these recommendations that are specific to each treatment group and with the goal of succinctly capturing clinical decision making. It should be noted that these disease activity levels are subjective and not strictly evidence based. However, the creation of discrete disease activity levels was requisite for the effective use of the RAND/UCLA Appropriateness Method.

These recommendations are based on 3 disease activity levels: low, moderate, and high. Patients with inactive disease were not considered. In a broad sense, the low disease activity level is meant to represent patients at the lowest disease activity level for which a majority of clinicians may consider altering the current medication regimen, and the high disease activity level is meant to represent patients with disease activity that is equivalent to or higher than the “average” subject that may have been enrolled in a clinical trial of the medications under consideration. The CEP considered baseline patient data from recently published clinical trials (10, 15, 17, 40-42), data from a recent attempt to define minimal disease activity in JIA (43), and their personal clinical experience in developing the disease activity levels for these recommendations. The disease activity levels are listed according to each JIA treatment group in Tables 1–5.

During the evaluation of the clinical scenarios, TFP members were allowed to recommend continuation of current therapy (not initiate a new therapeutic agent), regardless of the current disease activity level. In addition to a patient's current disease activity level, clinicians in practice may consider a patient's disease activity level prior to initiating the current treatment regimen when evaluating the effectiveness of the current treatment. That is, a current state of low disease activity may represent a recent major improvement for one patient and no improvement for another. Due to the complexity and number of clinical scenarios, previous disease activity levels were not included in the recommendation process and TFP members considered the appropriateness of initiating new therapeutic agents based on a patient's current disease activity level only.

Therapeutic agent definitions.

Medication classes used in the recommendation scenarios are defined as follows: NSAID refers to all nonsteroidal antiinflammatory drugs used commonly in clinical practice in the US and includes selective cyclooxygenase 2 inhibitors but not aspirin. Calcineurin inhibitors refer to cyclosporine and tacrolimus. Tumor necrosis factor α (TNFα) inhibitors refer to adalimumab, etanercept, and infliximab.

Clinical evaluation of the effectiveness of a given therapeutic agent may depend on the dose received. For all recommendations, it is assumed that patients have received the maximum tolerated typical dose of prior medications, as listed in Supplementary Appendix B (available in the online version of this article at http://onlinelibrary.wiley.com.bibliotheek.ehb.be/journal/10.1002/(ISSN)2151-4658). Of note, the dose of methotrexate was assumed to be 15 mg/m² (0.6 mg/kg) and administered via the parenteral route. These specific doses were used as a guide for TFP members when considering the scenarios; higher doses of these therapeutic agents may be appropriate in some clinical situations.

Clinical evaluation of the effectiveness of a given therapeutic agent may depend on the duration of therapy. Some scenarios explicitly stated the duration of the current treatment regimen. In scenarios where duration of therapy was not defined, it was assumed that the duration of therapy was sufficiently long to assess the response to therapy.

The recommendations assume that all medication regimens may contain a single NSAID as adjunct therapy when considering the appropriateness of initiating new therapies. For example, a patient described as currently receiving methotrexate is assumed to be taking methotrexate with concurrent use of a single NSAID, as appropriate.

Many of the recommendations refer to “initiating” a new therapy for patients presently undergoing treatment for JIA. In these cases, “initiating” is defined as either adding the new therapy while continuing current therapy or switching to the new therapy while discontinuing the current therapy. To limit the total number of scenarios, we made no distinction between adding and switching for some of the recommendations. Of note, combination biologic DMARD therapy was not considered by these recommendations, owing to concern from the reported increased incidence of infections in studies of rheumatoid arthritis in adult populations (44, 45). Accordingly, initiation of a new biologic DMARD is always intended to be accompanied by discontinuation of any current biologic DMARD.

Safety monitoring.

Scenarios were created to evaluate the appropriateness of several safety monitoring interventions for selected therapeutic agents and were voted upon by the TFP in the same manner as the treatment scenarios.

RAND/UCLA Appropriateness Method voting and scoring.

A teleconference was conducted to orient all TFP members to the voting task. Individual scenarios contained an explicit question that incorporated the key clinical parameters described above (e.g., “Rate the appropriateness of initiating TNFα inhibitor in a patient with a history of arthritis of 5 or more joints who has received methotrexate for 6 months and has moderate disease activity and features of poor prognosis”). There were 1,539 such individual scenarios and each TFP member evaluated all of them.

Each scenario was evaluated for “appropriateness,” which is defined as “the health benefits exceed the health risks by a sufficiently wide margin that the intervention is worth doing” (18). In this evaluation, the risks of initiating a particular therapeutic agent include the risk of not initiating an alternative agent (e.g., the risk of initiating NSAID monotherapy includes the risk of not initiating methotrexate with its greater potential for therapeutic benefit). The TFP scored each scenario on a 9-point scale, with scores of 7–9 denoting “appropriate,” 1–3 denoting “inappropriate,” and 4–6 denoting “uncertain,” i.e., either the risks and benefits are approximately equal or not enough information is available to make a meaningful evaluation. TFP members cast their votes while taking into account the currently available published literature (including but not limited to that provided to them in the evidence report) and their own personal clinical experience. Votes were cast independently by entering them into a formatted electronic spread sheet. The voting sheets were then electronically collected and combined to analyze the results.

The votes were analyzed in accordance with the RAND/UCLA Appropriateness Method (18), taking into account both the distribution and the median value of the votes. The scenario votes were first evaluated for disagreement (wide dispersion of votes), defined as one-third or more of the TFP assigning a score of inappropriate (1, 2, or 3), while one-third or more of the TFP assigned the same scenario a score of appropriate (7, 8, or 9). In the absence of disagreement, a median score of 3 or less classified a scenario as “inappropriate,” and a median rating of 7 or greater classified a scenario as “appropriate.” Those scenarios with median ratings in the 3.5 to 6.5 range, together with those where disagreement occurred, were classified as “uncertain.”

Following analysis of the first round of voting, a subsequent face-to-face meeting of the TFP was conducted and discussion was focused on scenarios where the voting outcome was “uncertain.” Following sufficient discussion on each group of scenarios, the TFP members voted for the final time. No attempts were made to “force” consensus among the TFP. Rather, the RAND/UCLA Appropriateness Method largely seeks to identify existing consensus (18).

Developing recommendations from votes.

Statements listed as recommendations met the RAND/UCLA Appropriateness Method criteria for “appropriate” in the final TFP votes with no wide dispersion of votes and a median score in the appropriate range (7 to 9). In some instances, more than one individual therapeutic agent was recommended for a given combination of clinical parameters; these are noted as being “recommended as one treatment approach.” Indications for initiating therapeutic agents that did not meet the criteria for appropriate were summarized as either uncertain or inappropriate. In these summary statements, uncertain means that some or all of the scenarios met the definition of “uncertain” and some or none of the scenarios met the definition of “inappropriate.” Inappropriate means that all of the scenarios met the definition of “inappropriate” and none met the definition of “uncertain.” The recommendations were reviewed in detail by all CEP members to ensure proper and clear translation of the TFP votes into text.

During the face-to-face discussion of the scenarios, several points were raised that the TFP members believed should be mentioned in the recommendations, even though they were not addressed in the voting scenarios. Following discussion, consensus on these points was reached through a simple show of hands vote. These statements are clearly noted in the results, are not considered formal recommendations, and were not assigned a level of evidence.

Rating evidence for recommendations.

For each final recommendation of appropriate or inappropriate, a level of evidence was assigned based on the methods of the University of Oxford Centre for Evidence-Based Medicine, UK (46). This categorization of evidence was used in the recently published European League Against Rheumatism recommendations for the management of rheumatoid arthritis (47) and incorporates the concept of “extrapolations,” where published data are used in a situation that has potentially clinically important differences from the original study situation. Such situations are frequently encountered when the patient's clinical factors in a recommendation do not match exactly with those of the supporting studies (19). The Oxford system also distinguishes between recommendations for which there is low-grade evidence and those for which there is an absence of published evidence.

Level of evidence “A” was assigned when the recommendation was supported by randomized clinical trials (or consistent inception cohort studies for questions of prognosis). Level of evidence “B” was assigned when the recommendation was supported by nonrandomized controlled studies (e.g., cohort and case–control studies) or extrapolations from randomized clinical trials. Level of evidence “C” was assigned when the recommendation was supported by uncontrolled studies (case series), extrapolations from nonrandomized controlled studies, or marked extrapolations from randomized clinical trials (e.g., studies of adult arthritis patients applied to juvenile arthritis or studies of polyarthritis phenotype applied to oligoarthritis). Level of evidence “D” was assigned when the recommendation was based on expert opinion without supporting published evidence.

Managing perceived potential conflicts of interest.

Perceived potential conflicts of interest were managed in a prospective and structured manner. All members of the CEP and TFP completed and submitted the ACR disclosure of interest form prior to participation in the project. The forms were updated prior to the face-to-face meeting. A summary listing of all perceived potential conflicts of interest was distributed to all project participants and was submitted for publication along with these recommendations.

Peer review of recommendations.

Following submission of the draft recommendations, the ACR invited peer review by members of the ACR Practice Guidelines Subcommittee, the ACR Quality of Care Committee, and the ACR Board of Directors. The recommendations were ultimately subject to the regular review process of this journal.

Updates to ACR recommendations.

We expect that knowledge of the appropriate therapeutic management of JIA will continue to advance. Clearly, these ACR recommendations will require updating to remain accurate and relevant. We suggest that approximately 3 years after the initiation of this current project (i.e., in 2012), the ACR Practice Guidelines Subcommittee and the ACR Quality of Care Committee evaluate newly published evidence and solicit expert opinion on the need for revision of these recommendations. Targeted updates focused only on areas of new published evidence may be appropriate.

Future research in the treatment of JIA may reasonably be directed by knowledge deficiencies identified by these recommendations. Of particular note to the CEP was the need for comparative long-term studies of the benefits and risks of early initiation of biologic therapies, development and implementation of continuous disease activity scores in clinical practice, studies of the evaluation and management of enthesitis, studies of the management of systemic arthritis and macrophage activation syndrome, and studies of the appropriate management of patients with inactive disease while receiving therapy.

General medication usage.

The TFP considered the general usage of intraarticular glucocorticoid injections and the use of methotrexate when initiating TNFα inhibitors. These recommendations apply broadly across treatment groups.

Initiation of therapeutic agents.

The appropriateness of initiating various therapeutic agents in the treatment of JIA is organized by treatment group. Each recommendation is characterized by the patient's clinical factors: treatment group, current medication, disease activity, and features of poor prognosis. Individual therapeutic agents are listed in the order of escalation of therapy as determined by the TFP. In situations where two or more agents were recommended for similar patient clinical factors, the agents are listed alphabetically. The recommendations are not mutually exclusive, i.e., the initiation of more than one therapeutic agent may be appropriate for a given set of patient clinical factors due to overlapping indications. Therapeutic agents that were considered by the TFP but not recommended for initiation are listed at the end of each treatment group in alphabetical order.

History of arthritis of 4 or fewer joints.

The definitions of disease activity and features of poor prognosis for this treatment group are listed in Table 1. A diagram of the overall treatment strategy is shown in Figure 1.

History of arthritis of 5 or more joints.

The definitions of disease activity and features of poor prognosis for this treatment group are listed in Table 2. A diagram of the overall treatment strategy is shown in Figure 2.

Active sacroiliac arthritis.

As stated in Materials and Methods, active sacroiliac arthritis was defined by the presence of clinical and imaging evidence. The definitions of disease activity and features of poor prognosis for this treatment group are listed in Table 3. The only medication class evaluated by the TFP for this treatment group was the TNFα inhibitors. There is no corresponding figure for this treatment group.

Systemic arthritis.

As noted in Materials and Methods, the JIA category of systemic arthritis was divided into two treatment groups: active systemic features and active arthritis. The appropriate treatment of patients with concurrent active systemic features and active arthritis may be expected to incorporate elements of both sets of recommendations. Given the variable course of systemic arthritis and the extent to which the prior disease course may influence subsequent treatment decisions, the TFP evaluated the appropriateness of initiating therapies for recently diagnosed patients, as opposed to patients with extensive and potentially influential disease histories.

As noted in Materials and Methods, the initiation of treatment with interleukin-6 inhibitors (such as tocilizumab) or interleukin-1 inhibitors other than anakinra (such as canakinumab or rilonacept) was not considered in the development of these recommendations because these therapeutic agents were not widely commercially available at the time. Options for the treatment of systemic arthritis appear to be increasing. The appropriateness of initiating recently available therapeutic agents for the treatment of systemic arthritis may need to be the focus of a timely update to these recommendations.

Systemic arthritis with active systemic features (and without active arthritis).

The definitions of disease activity and features of poor prognosis for this treatment group are listed in Table 4. A diagram of the overall treatment strategy is shown in Figure 3.

These recommendations are not meant to apply to patients with clinical and laboratory evidence of macrophage activation syndrome that warrants specific modification of therapy. Life-threatening clinical scenarios (e.g., cardiac tamponade) were not considered by the TFP and may warrant deviation from these recommendations. TNFα inhibitors were not considered by the TFP in the treatment of active systemic features, owing to their reported relatively poor effectiveness (12, 89, 95-98).

Systemic arthritis with active arthritis (and without active systemic features).

The definitions of disease activity and features of poor prognosis for this treatment group are listed in Table 5. A diagram of the overall treatment strategy is shown in Figure 4. Consistent with the assessment of other treatment groups, the appropriateness of systemic glucocorticoids for the treatment of arthritis in patients with systemic arthritis was not addressed.

Safety monitoring.

The TFP evaluated the appropriateness of several safety monitoring interventions by considering the health benefits (minimizing the chance of toxicity or adverse events) and harms (frequent phlebotomies, false-positive test results). As with the evaluation of initiation of therapeutic agents, the TFP did not consider the economic costs of monitoring or adverse events. A summary of the recommendations for medication safety monitoring is shown in Table 6.

Addendum.

Therapies that were approved after the original literature review are not included in these recommendations.