Comparative Effectiveness of Combining MTX with Biologic Drug Therapy Versus Either MTX or Biologics Alone for Early Rheumatoid Arthritis in Adults: a Systematic Review and Network Meta-analysis.

Background Comparative effectiveness of early rheumatoid arthritis (RA) treatments remains uncertain. Purpose Compare benefits and harms of biologic drug therapies for adults with early RA within 1 year of diagnosis. Data Sources English language articles from the 2012 review to October 2017 identified through MEDLINE, Cochrane Library and International Pharmaceutical Abstracts, gray literature, expert recommendations, reference lists of published literature, and supplemental evidence data requests. Study Selection Two persons independently selected studies based on predefined inclusion criteria. Data Extraction One reviewer extracted data; a second reviewer checked accuracy. Two independent reviewers assigned risk of bias ratings. Data Synthesis We identified 22 eligible studies with 9934 participants. Combination therapy with tumor necrosis factor (TNF) or non-TNF biologics plus methotrexate (MTX) improved disease control, remission, and functional capacity compared with monotherapy of either MTX or a biologic. Network meta-analyses found higher ACR50 response (50% improvement) for combination therapy of biologic plus MTX than for MTX monotherapy (relative risk range 1.20 [95% confidence interval (CI), 1.04 to 1.38] to 1.57 [95% CI, 1.30 to 1.88]). No significant differences emerged between treatment discontinuation rates because of adverse events or serious adverse events. Subgroup data (disease activity, prior therapy, demographics, serious conditions) were limited. Limitations Trials enrolled almost exclusively selected populations with high disease activity. Network meta-analyses were derived from indirect comparisons relative to MTX due to the dearth of head-to-head studies comparing interventions. No eligible data on biosimilars were found. Conclusions Qualitative and network meta-analyses suggest that the combination of MTX with TNF or non-TNF biologics reduces disease activity and improves remission when compared with MTX monotherapy. Overall adverse event and discontinuation rates were similar between treatment groups. Registration PROSPERO (available at http://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD42017079260).


INTRODUCTION
Rheumatoid arthritis (RA) is an autoimmune systemic inflammatory disease affecting more than 1 million Americans and characterized by synovial inflammation, which can lead to progressive bone erosion, joint damage, and disability. 1 For patients with early RA (≤ 1 year of disease), 2 guidelines recommend early treatment with the goal of remission or low disease activity. 3,4 Available therapies for RA include corticosteroids, conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), tumor necrosis factor (TNF) and non-TNF biologics, targeted synthetic DMARDs (tsDMARDs), and biosimilars. Over the past two decades, biologics have become an important treatment option for established RA. However, clinicians face the challenge around biologic use in early RA.
Biologics commonly used for RA treatment include TNF biologics (adalimumab, certolizumab pegol, etanercept, golimumab, infliximab) and non-TNF biologics (abatacept, rituximab, tocilizumab, and sarilumab). Experts and guideline groups support using csDMARDs, often methotrexate (MTX), as the first line-therapy. 3,4 Despite recommendations, advocates encourage early biologic use to induce remission. 5,6 In a 2012 systematic review, evidence comparing early RA treatment options was limited. 7 No studies investigated efficacy, effectiveness, and harms among subgroup populations. Recently, information from clinical trials of four biosimilar drugs (ADA-atto, IFX-dyyb, IFX-abda, ETNszzs), a tsDMARD (tofacitinib), and one non-TNF biologic (sarilumab) have become available. Additionally, studies continue to be published on established therapies. Given this uncertainty, the Agency of Healthcare Research and Quality (AHRQ) and the Patient Centered Outcomes Research Institute (PCORI) commissioned a systematic review to compare effectiveness and harms of RA drugs in patients with early RA. This paper focuses on comparisons of benefits and harms of treatments in early RA involving biologics.

METHODS
The full technical report describes the study methods in detail 8 and the protocol was registered at PROSPERO (http://www. crd.york.ac.uk/PROSPERO/display_record.php? ID=CRD42017079260). In a comprehensive synthesis of the evidence, we included data from studies dating back to June 2006, identified in the 2012 review on this topic 7 and through an updated literature search.

Data Sources
A professional research librarian searched MEDLINE, Cochrane Library, Embase, and International Pharmaceutical Abstracts from January 2011 to October 5, 2017. We rereviewed studies included in the 2012 review, 7 supplemental evidence (data received through the AHRQ Web site and a Federal Register notice), and reference lists of included studies and recent reviews. We also searched the following sources for unpublished studies: ClinicalTrials.gov, World Health Organization International Clinical Trials Registry Platform, and New York Academy of Medicine's Grey Literature Index.

Study Selection
We included study populations defined as early RA by the authors if the diagnosis was ≤ 1 year in the past (Table 1 presents inclusion and exclusion criteria). Two reviewers independently reviewed titles and abstracts using abstrackr 9 and full-text articles for eligibility. To assess efficacy regarding disease activity, response, remission, radiographic progression, and functional capacity, we included head-to-head controlled trials and prospective cohort studies comparing any of the therapies. In addition, we included placebo-and MTXcontrolled trials for network meta-analyses (NWMA). For adverse events, we abstracted data on overall adverse events, overall study discontinuation, discontinuation attributed to adverse events or toxicity, patient adherence, and any serious adverse events as defined by the FDA. 10 For specific adverse events (that were not serious adverse events), we focused on those most commonly reported according to their FDAapproved labels.

Data Extraction and Risk of Bias Assessment
Trained reviewers abstracted each study using a structured, pilot-tested form and a senior reviewer evaluated accuracy. To assess the risk of bias (ROB), we adapted the Cochrane Risk of Bias tool 11 for randomized controlled trials (RCTs) and used the Risk of Bias in Non-randomised Studies of Interventions (ROBINS-I) tool 12 for nonrandomized controlled studies.

Data Synthesis and Analyses
We planned to conduct pairwise analysis when possible and NWMAs to estimate the indirect treatment effects. Criteria for eligible studies for NWMA included 1 no failed prior treatment attempt with MTX 2 , treatment doses within FDA-approved ranges 3 , 12-month follow-up, and 4 double-blinded RCTs of low or medium ROB. Head-to-head and placebo-controlled RCTs were eligible for NWMA; however, we did not find any eligible placebo-controlled trials in a population with early RA. We considered NWMA for American College of Rheumatology 50% improvement (ACR50), Disease Activity Score (DAS) remission, radiographic joint damage, all study discontinuations, and discontinuations attributed to adverse events.
We ran NWMAs using a multivariate, random effects metaregression model with restricted maximum likelihood for variance estimation. 13 Models were fit using the Stata "network" package 14 an updated versions of the "mymeta" package which accounts for multi-arm trials. The network structure for outcomes was mostly "star-shaped," indicating a dearth of head-tohead studies directly comparing interventions (see Figs. 1 and 2, low strength of evidence). Most effect estimates, therefore, were derived from indirect comparisons relative to MTX rather than mixed treatment comparisons. For closed loops, we tested the transitivity assumption by examining loop-specific consistency between direct and indirect effects using network side splits and global consistency by comparing a model assuming consistency with a model not assuming consistency (i.e., inconsistency model). When the global Wald test indicated no significant differences between the consistency and inconsistency models, 15 and no significant differences in estimates based on side splits, we presented consistency model estimates.

Strength of Evidence
We evaluated strength of evidence for each comparison based on the guidance established for AHRQ's EPC Program as high, moderate, and low or insufficient. 16 We graded strength of evidence for the following outcomes: disease activity, response, radiographic joint damage, functional capacity, discontinuation because of adverse events, and serious adverse events.

Role of the Funding Source
This topic was nominated and funded by PCORI in partnership with AHRQ. The AHRQ Contracting Officer's Representative and PCORI program officers provided comments on the protocol and full evidence report. Neither PCORI nor : JGIM AHRQ directly participated in literature searches; study eligibility criteria determination; data analysis or interpretation; or preparation, review, or manuscript approval for publication.

Characteristics of Reviewed Studies
We identified 6373 citations from electronic searches and 429 from other sources (Fig. 3). We were unable to use pairwise meta-analyses due to a lack of head-to-head studies. In this paper, we report results from trials of biologic comparisons only. For these comparisons, we found 22 RCTs with low or medium risk of bias (Table 2). We included 13 studies in our NWMA.
Study durations ranged from 6 months to 2 years. Over half of the study populations were women (range 53 to 81%) with mean ages ranging from 46 to 57 years. Included studies almost exclusively enrolled patients with high disease activity at baseline as measured by mean or median Disease Activity Score (DAS) 28 scores (range of mean scores 3.6 to 7.1).
Among studies reporting MTX use, 18 studies (82%) enrolled MTX-nave patient samples; the remaining 3 studies enrolled patients with prior csDMARD use (including MTX). Most trials used ACR response, disease activity scores to measure clinical improvement, and Sharp or Sharp/van der Heijde scores to measure radiographic progression of the disease. Trials examining function or quality of life most commonly used the Health Assessment Questionnaire (HAQ) or Medical Outcomes Study Short Form 36 (SF-36). Harms studies generally described overall withdrawals, withdrawals due to adverse events, and specific adverse events including the most commonly occurring across all eligible drugs according to their FDA-approved labels. The majority (N = 21, 95%) were at least partially industry funded. Table 3 summarizes main findings and the strength of evidence. The remainder of results are organized such that we present evidence on the combination of biologics with MTX compared first to biologic monotherapy and second to MTX monotherapy for both TNF and non-TNF biologics. NWMA, when available, follows the comparisons.  : JGIM plus MTX had smaller radiographic changes and higher remission rates than MTX monotherapy. NWMA found lower overall discontinuation rates for combination therapy consisting of TNF biologics (specifically, ADA, CZP, and ETN, but not IFX) plus MTX than MTX monotherapy (range of RR, 0.64 [95% CI, 0.53 to 0.78] to 0.66 [95% CI, 0.43 to 1.00]) (data not shown). However, neither serious adverse events nor discontinuations because of adverse events differed between the groups.
Infliximab. Three trials examined the combination of IFX plus MTX compared with MTX monotherapy in MTX-nave patients. [45][46][47] The largest trial (n = 1049) compared the efficacy of initiating two different combinations of IFX (3 mg/kg or 6 mg/kg) plus MTX (20 mg/week) with MTX monotherapy over 54 weeks 45 49 and less radiographic progression (modified SHS change 0.4 to 0.5 for IFX combination therapy groups vs. 3.7 for MTX only, p < 0.001). 45 The smaller trials found improved 46 or trending 47 ACR50 responses in favor of IFX combination therapy at 54 weeks among patients receiving IFX plus MTX combination therapy.

Non-TNF Biologic Plus Methotrexate Versus Non-TNF Biologic Monotherapy. Abatacept (ABA).
One RCT, the multinational AVERT trial (n = 351), compared the combination of ABA (125 mg/week) plus MTX (7.5 mg/week) with ABA monotherapy. 50 This double-blind RCT compared treatments over 1 year; at year 2, patients with DAS28-CRP < 3.2 were tapered off treatment. If patients experienced an RA flare by month 15, they were given ABA plus MTX. No significant differences were noted for ABA plus MTX versus ABA monotherapy for ACR50 response, remission (DAS28-CRP < 2.6), or functional capacity.

Non-TNF Biologic Plus
The NWMA found significant differences in ACR50 response when comparing ABA plus MTX with MTX monotherapy (RR, 1.34; 95% CI, 1.16 to 1.54), consistent with results from the AGREE and AVERT trials (Fig. 4). The combination of ABA plus MTX had numerically less radiographic progression than MTX monotherapy, but the difference was not significant (SMD, − 0.09; 95% CI, − 0.26 to 0.09) (Fig. 5).
In NWMA, there was no difference in overall discontinuation between ABA plus MTX and MTX alone, though ABA plus MTX had fewer discontinuations due to adverse events (RR, 0.49, 95% CI, 0.28 to 0.86) (data not shown).
Tocilizumab. Two RCTs, the FUNCTION trial 51 (N = 1162) and the U-Act-Early trial 52 (N = 317), both previously described in the "Non-TNF Biologic Plus Methotrexate Versus Non-TNF Biologic Monotherapy" section, assessed differences in efficacy between TCZ plus MTX and MTX monotherapy in MTX-nave populations. In both trials, TCZ plus MTX combination therapy led to smaller radiographic changes and higher remission rates (DAS28-ESR < 2.6: 49% vs. 19.5%, p < 0.001) than MTX monotherapy after 1 to 2 years. Both trials demonstrated greater functional capacity in the combination group than the MTX monotherapy group. Overall discontinuation rates, discontinuation because of adverse events, and serious adverse events did not differ across groups.

Subgroups
Only three RCTs compared drug therapies among different subgroups defined by demographics, disease activity, or coexisting conditions. 33, 38, 45 We could not draw any conclusions about response rates or serious adverse events between older and younger patients or between people with different levels of disease activity.

DISCUSSION
Although several biologic agents are available, head-to-head evidence remains limited. Combination therapy with TNF or non-TNF biologics plus MTX resulted in improved disease control, DAS-defined remission, and functional capacity compared with monotherapy of either MTX or a biologic. Network meta-analyses (NWMAs) found higher ACR50 response for combination therapy of biologic plus MTX than MTX monotherapy. The results of comparative NWMA for overall discontinuation and discontinuation attributed to adverse events had confidence intervals too wide to support firm conclusions. Subgroup data were limited. Eligible early RA studies almost exclusively included patients with high disease activity. In contrast, patients with early RA may present in a clinical setting with varying levels of severity. Patients with early RA enrolled in trials consist of highly selected individuals. 60 Although the evidence for the effectiveness of biologics plus MTX in early RA is favorable, it is not the standard of care for several reasons. First, some data indicate that certain patients will do well on MTX monotherapy, but no information is available about how to identify Figure 4 Forest plot for network meta-analysis (low SOE grades for all NWMA effect estimates) of biologic plus MTX vs. biologic or MTX only: ACR50 response rates. 95% CI, 95% confidence interval; MTX, methotrexate; RR, relative risk; TNF, tumor necrosis factor; vs., versus.
: JGIM or predict who these patients will be. 33,38,45 Second, many insurers require MTX failure as a prerequisite to add a biologic. Third, some patients may be wary of a combination therapy approach in early disease (e.g., cost, side effects, injections).
Several limitations of our review should be considered. No consensus exists on defining early RA. For this review, we defined populations with early RA as having a diagnosed disease duration of 1 year or less and included mixed population studies if > 50% of the study populations had an early RA diagnosis. Patients described in this way may have had longer symptoms. Although existing evidence of biologics in combination with MTX shows that this regimen can improve disease activity, we do not know whether starting biologic treatment rather than MTX improves long-term prognosis. Because of a lack of head-to-head trials, we often relied on NWMA to estimate the comparative effectiveness of interventions of interest for treating patients with early RA. NWMAs are an important analytic tool in the absence of direct head-to-head evidence, but also have limitations; thus, we graded them as low strength of evidence. NWMAs often yield estimates with wide confidence intervals that encompass clinically relevant benefits or harms for both drugs (or combination therapies) being compared. The network geometry was mostly starshaped with very few closed loops, which limited the number of tests we could use to assess transitivity and consistency. The FDA has approved several biosimilars, but there were no eligible studies of biosimilars.
Future research directions include comparisons of patients with different degrees of disease activity or poor prognostic factors and longer-term effects. Data are needed for examination of biosimilars. Studies with longer treatment periods and follow-up of 5 years or longer would provide better information on long-term effectiveness, adherence, and adverse events. They would also yield insights as to whether starting with a biologic improves long-term prognosis of RA.