The RAndomized Placebo Phase Study Of Rilonacept in the Treatment of Systemic Juvenile Idiopathic Arthritis (RAPPORT)

Patients

The study was conducted in compliance with principles of the International Conference on Harmonization and Good Clinical Practice. The study was approved by the Institutional Review Boards of each study site. All patients or parents/guardians provided written informed consent. An independent Data Safety Monitoring Board, appointed by NIH/NIAMS, met every 6 months to evaluate study conduct and safety. Twenty Childhood Arthritis and Rheumatology Research Alliance (CARRA) centers in the US enrolled patients from 11/2008 to 5/2012. Key Inclusion criteria included: International League against Rheumatism criteria for sJIA ¹⁸; age ≥18 months to ≤19 years; ≥2 active joints; stable methotrexate dose for ≥4 weeks; stable corticosteroids ≥2 weeks; ≤ 2mg/kg or 60mg prednisone or equivalent. If previously treated with biologics, the following lengths of discontinuation were required: anakinra ≥4 days, etanercept ≥4 weeks, adalimumab ≥6 weeks, tocilizumab ≥6 weeks, abatacept ≥6 weeks, and infliximab ≥8 weeks. Key exclusion criteria included current treatment with disease modifying anti-rheumatic drug other than methotrexate; intra-articular corticosteroids or pulse steroids within 4 weeks; leflunomide without cholestyramine wash out; cyclophosphamide within 3 months; IVIG within 4 weeks; treatment in the past with an IL-1 inhibitor other than anakinra; renal insufficiency as defined as an elevated serum creatinine; aspartate aminotransferase (AST) or alanine aminotransferase (ALT) > 2 times the upper limit of normal; thrombo-, leuko-, or neutro-penia; prolonged PT or PTT; positive PPD without treatment documentation; live virus vaccine within 1 month; pregnancy or sexual activity without contraception.

Study Design

RAPPORT incorporated an initial 4-week double-blind placebo phase within a 24-week randomized multi-center design. As such, patients were randomized 1:1 to treatment with (a) 4 weeks of placebo followed by 20 weeks of rilonacept or (b) 24 weeks of rilonacept, resulting in a double-blind phase (weeks 0-4) and an all active treatment phase (weeks 4-24). This was followed by an open label Long Term Extension (LTE) phase (24 weeks-21 months). The randomized placebo phase study design is based on the assumption that if treatment is effective, patients who receive active drug earlier will respond sooner, on average, than patients who receive active drug later ¹⁹. This design is especially useful when testing highly effective therapies with rapid onset of action of several weeks, and when minimizing time on placebo or safety issues are important ¹⁹. IL-1 inhibition appears to have an onset of action of 2 weeks, based on data in sJIA subjects in the Amgen-sponsored study of anakinra in polyarticular-JIA that included subjects with sJIA ¹². Randomization was performed using a Web-based randomization and drug supply management system (WebEZ, Almac, Durham, North Carolina). The central randomization scheme used a fixed-block size, stratified according to baseline corticosteroid use. Rilonacept and matching placebo were provided by Regeneron Pharmaceuticals, Inc. (New York).

In the rilonacept arm, a loading dose (4.4mg/kg, maximum dose 320mg) of rilonacept was given on day 0 followed by weekly maintenance doses (2.2mg/kg, to a maximum dose 160mg). In the placebo arm, a loading dose of placebo was given on day 0 followed by 3 maintenance doses of placebo; then a loading dose of rilonacept was given on day 28 followed by weekly maintenance doses of rilonacept. Evaluation of efficacy occurred during the first 12 weeks of the study; safety was assessed in the first 24 weeks and in the LTE. All patients who benefited from rilonacept treatment as judged by the treating physician were eligible for enrollment in the LTE phase. Patients in the LTE were initially offered open label rilonacept for 2 years or until rilonacept was commercially available. These criteria were shortened for some patients because of budgetary issues.

Clinical Assessments

Screening visits could occur up to 4 weeks before randomization. A randomization/baseline visit occurred at week 0 and follow up visits occurred at weeks 2, 4, 6, 8, 10, 12, 14, 18 (by telephone), and 24. Patients were seen every 3 months in the LTE. The 2 week interval between visits in the first 14 weeks was used to maximize the precision of estimation of time to response. Medical history, physical examination, concomitant medications, adverse events, fever (patients’ daily diary), blinded joint assessment, physician global assessment, parent/guardian/patient global assessment, clinical laboratory tests, and Childhood Health Assessment Questionnaire (CHAQ) were assessed at every (non-telephone) visit. Fasting lipoprotein profiles and PedsQL questionnaires were assessed at week 0, 12 and 24. Biospecimens (DNA, RNA, whole blood, serum, plasma) were obtained during the trial.

Endpoints

The primary endpoint was time to response during the 12 week efficacy period. Response was defined as a composite of 1)improvement in the JIA ACR30 ²⁰, 2) absence of fever ≥38.5°C in the previous 2 weeks, and 3)at least 10% taper in systemic corticosteroids from baseline if the patient were taking corticosteroids. The JIA ACR30 algorithm required at least 30% improvement from baseline in at least 3 of 6 core variables, with no more than 1 variable worsening by 30% or more. Corticosteroids were required to be tapered if all of the following criteria developed by consensus among the RAPPORT investigators ²¹ were met: fever ≤2 days in previous 7 days, absence of poor physical function, and absence of laboratory values associated with impending MAS. The time to response was defined as the visit designation at which the patient first achieved the response criteria and maintained response until the next scheduled visit. Absence of rash was not included in the response criteria.

Corticosteroids were increased or started if one or more of the following criteria developed by consensus of the RAPPORT investigators were met²¹: MAS ²², incomplete MAS, symptomatic anemia with a hemoglobin ≤6.5g/dL, myocarditis, or symptomatic pneumonitis, or serositis unresponsive to NSAIDs. Patients who met criteria for corticosteroid increase/start were deemed non-responders. Patients who met the criteria for non-response and started/increased corticosteroids were no longer eligible for response.

Secondary endpoints included the JIA ACR30, 50, 70, inactive disease ²³, presence of fever, serositis, symptomatic anemia, abnormal liver function, rash, MAS, incomplete MAS, corticosteroid dose, CHAQ, and PedsQL Generic Core Modules. Adverse events (AEs) and Serious Adverse Events (SAEs) were collected throughout the study.

Data Analysis

Planned enrollment was 100 patients (50 per arm) to achieve ≥80% power to detect a difference at a 2-sided α level of 0.05. This sample size was determined based on the results from multiple Monte-Carlo simulations conducted under various assumptions for distribution of time to response in the two study arms.The trial was ended before reaching the enrollment goal due to slow enrollment and financial considerations. One pre-specified interim analysis of the primary endpoint to examine early stopping for overwhelming efficacy was performed when data were available for 50 patients. The Lan-DeMets flexible spending function corresponding to the O'Brian-Fleming stopping boundary was used to preserve the overall type I error of 0.05²⁴. The stopping boundary P value at the interim analysis was 0.007. After reviewing the efficacy data at the interim analysis, the DSMB recommended to continue the study. The 2-sided P value for rejecting the null hypothesis of no difference between the treatment arms for the primary endpoint at the final analysis was 0.048. A 2-sided α level of 0.05 was considered statistically significant for all other endpoints. P values for the secondary analyses were not adjusted for multiple comparisons. All analyses were performed in SAS version 9.2 (SAS Institute Inc., Cary, NC).

Data were analyzed on an intent-to-treat basis. Descriptive statistics were summarized by treatment arm. The primary endpoint analysis compared time to response between treatment arms using Gehan-Wilcoxon test, which emphasizes early differences. Missing temperatures were imputed as fever free when calculating the response endpoint. A sensitivity analysis was performed treating the response endpoint as missing if the fever criterion could not be determined due to missing temperatures. In both the primary and sensitivity analyses, patients were censored if the response endpoint was missing at two consecutive visits, or if the patient met criteria for non-response and started/ increased corticosteroids prior to meeting the response criteria. Response rates were calculated using cumulative incidence estimation, treating non-response as a competing event. Comparison of the response to rilonacept vs. placebo at the end of the placebo phase (week 4) was evaluated using Fisher's exact test. Logistic regression was used to adjust for sJIA duration and presence of articular without systemic symptoms and to explore the association between those baseline characteristics and response at week 4.

JIA ACR30, 50, and corticosteroid dose were analyzed using a generalized estimating equation (GEE) repeated measures model. Improvement in JIA ACR70 was analyzed using Fisher's exact test. Adverse events were compared using chi-square test. Infection rates were compared using Poisson regression.

Study Population

Seventy-one patients were randomized (Figure 1). Baseline demographic and disease characteristics were similar between the two arms of the study, except recent fever was more common in the rilonacept group (Table 1). Fourteen patients withdrew from the study prior to week 24: 8 for inadequate response (7 placebo arm, 1 rilonacept arm), 2 withdrew consent (1 in each arm), 1 for a serious adverse event (elevated liver transaminases) in the rilonacept arm, 1 excluded before drug exposure for not meeting inclusion/exclusion criteria, 2 lost to follow up (1 in each arm). Of the 57 completers, 40 enrolled in the LTE; 29 completed the LTE and 11 did not.

Efficacy

The primary endpoint, time to response as defined by time to achieve the composite endpoint of JIA ACR30, absence of fever, and corticosteroid tapering (for patients taking corticosteroids), was shorter in the rilonacept arm (median 4 weeks; 25^th, 75^th percentiles 2,10 weeks) than in the placebo arm (median 8 weeks; 25^th percentile 6 weeks, 75^th percentile not estimable from the data) (Figure 2; chi-square 7.235; P=0.007). By week 12, 27 of 35 patients (77%) receiving rilonacept continuously from onset of the trial and 20 of 34 patients (59%) receiving placebo for the initial 4 weeks, met the primary endpoint of response. The sensitivity analysis of time to response without fever imputation also demonstrated shorter time to response in the rilonacept arm as compared to the placebo arm (chi-square = 5.270; P=0.022). Secondary endpoint analysis of the response rate at 4 weeks showed 20 of 35 patients in the rilonacept arm compared to 9 of 33 patients in the placebo arm (57% vs. 27%; P=0.016). The JIA ACR30, 50, and 70 response rates were significantly better in the rilonacept arm at week 4 compared to the placebo arm (Table 2; all P<0.05). Twenty-six of 35 (74%) patients in the rilonacept arm compared to 13 of 33 (39%) in the placebo arm met JIA ACR30 response criteria at week 4 (OR=4.54; 95% CI 1.62, 12.72; P=0.004); 21 of 35 (60%) vs. 10 of 33 (30%) in the placebo arm met JIA ACR50 (OR=3.50; 95% CI 1.28, 9.56; P=0.015); 14 of 35 (40%) vs. 4 of 33 (12%) met JIA ACR70 (P=0.013).

A pre-specified logistic regression analysis adjusted for sJIA duration and presence of articular involvement without systemic symptoms showed that the significant rilonacept effect on response at week 4 persisted after adjustment (OR=3.42; 95% CI 1.21, 9.70; P=0.020). We did not observe a statistically significant difference in odds of response at week 4 for patients without systemic manifestations at baseline compared to patients with systemic manifestations (OR=0.87; 95% CI 0.32, 2.40; P=0.794). There was no statistically significant difference in odds of responding at week 4 as a function of longer sJIA duration (OR=0.91; 95% CI 0.75, 1.11; P=0.359). No statistically significant difference in response rates at week 4 was observed for 26 patients unexposed to anakinra (44%) compared to 24 patients previously exposed to anakinra (40%); 8 subjects had missing data regarding anakinra exposure; 2 had missing response data at week 4.

Seventeen of the 50 patients taking corticosteroids at baseline discontinued corticosteroids during the study. Overall, corticosteroid dose decreased more in the rilonacept arm than in the placebo arm during the efficacy period (P=0.036; Figure 3). Laboratory tests reflecting disease activity are reported by treatment arm in Table 2.

Safety

In the double-blind and treatment phases, the rilonacept arm performed at least as well as the placebo arm in terms of safety (Table 3). There was not a higher incidence of infection in the rilonacept arm in either phase. Four patients, all in the rilonacept arm, developed elevations in liver transaminases of ≥2 times the upper limit of normal; 2 patients developed elevations ≥5 times the upper limit of normal (one of these was considered an SAE). There were 14 SAEs: 9 among patients in the rilonacept arm and 5 in the placebo arm with the most common being sJIA flare (4). The AST liver function test was consistently higher in the rilonacept arm.