Long-term safety and effectiveness of etanercept in JIA: an 18-year experience from the BiKeR registry

Background At present, etanercept represents the most commonly prescribed biologic agent for juvenile idiopathic arthritis (JIA) treatment. Children and adolescents with JIA are often treated with etanercept over long periods, sometimes even into adulthood. The objectives of this analysis were to determine the long-term safety of etanercept compared to a biologic-naïve cohort and to assess the long-term treatment response upon continuous etanercept exposure using data from the German biologics registry (BiKeR). Methods JIA patients newly exposed to etanercept were documented in the BiKeR registry from January 2001 to March 2019, and baseline characteristics, effectiveness, and safety parameters were analysed. Response to treatment was assessed according to 10-joint Juvenile Arthritis Disease Activity Score (JADAS10), JADAS-defined minimal disease activity and remission, JIA-American College of Rheumatology (ACR) improvement criteria, and ACR-inactive disease definition. Safety assessments were based on adverse event (AE) reports. Results A total of 2725 new etanercept users with a diagnosis of JIA were registered. Of these, etanercept was received as a first-line biologic by 95.8% and as monotherapy without concomitant methotrexate by 31.5%. After nine years on continuous treatment, 68.1% of patients presented minimal disease activity, 43.1% JADAS-defined remission on drug, and 36.6% ACR-inactive disease. JIA-ACR30/50/70/90 response rates were still 82/79/71/54% after nine years of treatment. Overall, 2053 AEs (34.3/100PY), including 226 serious AEs (SAE, 3.8/100PY), were observed upon etanercept, compared to 1345 AEs [35.6/100PY; p = 0.3] and 52 SAEs (1.4/100PY; p = 0.0001) in the biologic-naïve cohort. Respective exposure-adjusted rates for etanercept and biologic-naïve patients were 0.9/100PY and 0.2/100PY (p = 0.0001) for serious infections, 0.4/100PY and 0.1/100PY (p = 0.01) for zoster reactivation, 0.3/100PY and 0.03/100PY (p = 0.015) for inflammatory bowel disease, and 1.9/100PY and 1.4/100PY (p = 0.09) for uveitis. Three and two malignancies were documented in the etanercept and biologic-naïve groups, as well as three and one deaths, respectively. Conclusions No new safety signal was observed, especially no increased risk for malignancies or autoimmune disorders other than inflammatory bowel disease. However, SAEs and serious infections, though infrequent, were more often reported on etanercept than in biologic-naïve patients. In addition, etanercept demonstrated a long-term maintenance of clinical benefits up to nine years of continuous treatment.

(Continued from previous page) Conclusions: No new safety signal was observed, especially no increased risk for malignancies or autoimmune disorders other than inflammatory bowel disease. However, SAEs and serious infections, though infrequent, were more often reported on etanercept than in biologic-naïve patients. In addition, etanercept demonstrated a long-term maintenance of clinical benefits up to nine years of continuous treatment.
Keywords: Juvenile Idiopathic Arthritis, JIA treatment, Etanercept, TNF inhibitors, Biologics registry, Drug surveillance Background Treatment of juvenile idiopathic arthritis (JIA) represents a major challenge in paediatric rheumatology. Diverse treatment options are currently available. Methotrexate is the most commonly prescribed conventional diseasemodifying anti-rheumatic drug (DMARD). Within biologics, etanercept, a tumour necrosis factor inhibitor (TNFi), was the first drug to be approved for JIA in 2001 and represents at present the favoured first-line biologic agent for JIA patients [1]. Etanercept is approved for use in polyarticular JIA in children older than 2 years of age and for use in psoriatic arthritis (PsA) and enthesitisrelated arthritis (ERA) in patients older than 12 years of age. For systemic JIA or persistent oligoarthritis, etanercept is not approved, so that its use in these conditions is mostly reserved for children who experience refractory disease [2]. From 2001 to present, the increasing use of etanercept in patients with JIA has raised awareness of rare serious adverse events, such as malignancies and autoimmune conditions, including, but not limited to, uveitis, inflammatory bowel disease, and demyelinating disorders [3][4][5]. Also, etanercept treatment may be required over many years in JIA patients, sometimes even into adulthood. Yet, knowledge about its safety and effectiveness in the long-term is limited.
The German registry for biologics in paediatric rheumatology (BiKeR) is one of the largest national registries on the use of biologics in JIA. Over a period of 18 years, it has accumulated a large quantity of data on etanercepttreated JIA patients. We performed a systematic review of the BiKeR registry to evaluate the long-time safety and effectiveness of etanercept in JIA. A biologic-naïve cohort was used as a comparator for long-term safety analyses.

Methods
The German BiKeR registry has been documenting treatment of JIA with biologics since 2001 and has been extensively described in previous reports [6,7]. It was approved by the ethics committee of the physician board Aerztekammer Nordrhein, Duesseldorf. The BiKeR registry is registered in the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP [8]). Written consent was obtained from patients and parents, and repeated when the patient became an adult. Pseudonymized data were collected for each JIA patient starting a biologic therapy and belonging to the seven ILAR-defined JIA categories [9] as determined by the reporting physician. Dose and frequency of administration were documented. Patient assessment regarding effectiveness and occurrence of adverse events (AEs) was performed at baseline and at follow-up after three and six months and every six months thereafter. After discontinuation of treatment, patients were followed up every six months with a request to report any AE, and patients transitioning to adult care are followed up by the JuMBO registry [10]. Patients of the registry newly starting treatment with etanercept from January 2001 to March 2019 were included in the study if they had assessments at baseline and at least at the three-month visit, irrespective of diagnosis. All follow-up forms received prior to April 2019 were evaluated. Reasons for discontinuation of etanercept treatment were also documented. Multiple reasons could be given. JIA patients who newly started methotrexate treatment and never received biologics were recruited between 2005 and 2011 till inclusion of 1500 patients and served as the control group for long-term safety analyses.

Assessment of effectiveness
Effectiveness parameters were defined as follows. The JIA-American College of Rheumatology (ACR) improvement criteria and the Juvenile Arthritis Disease Activity Score (JADAS) were calculated as previously described [11,12]. JIA-ACR core set parameters consist of (i) physician global assessment of disease activity (Phys-VAS) on a 10-cm visual analogue scale (VAS); (ii) parent/patient global assessment of overall well-being (PatVAS) on a 10-cm VAS; (iii) the Childhood Health Assessment Questionnaire (CHAQ); (iv) the number of joints with active arthritis, defined by the presence of swelling or, if no swelling is present, limitation of motion accompanied by pain, tenderness, or both; (v) the number of joints with limited range of motion; and (vi) the erythrocyte sedimentation rate (ESR). The ACRinactive disease definition was used according to Wallace et al. [13], requiring no active uveitis or arthritis, no fever, rash, splenomegaly, serositis, generalised lymphadenopathy or elevation of ESR/C-reactive protein (CRP), best possible PhysVAS, and duration of morning stiffness ≤ 15 min. JADAS10 was chosen, which considers a maximum of ten active joints besides PatVAS, PhysVAS, and ESR or CRP, all equally weighted. Rates of JADAS-minimal disease activity (MDA) and JADASremission, respectively defined as JADAS10 ≤ 3.8 and JADAS10 ≤ 1, were calculated according to the definition of Consolaro et al. [14]. For each timepoint, all patients with complete data set were considered. Data from patients who discontinued etanercept treatment due to remission were analysed for disease activity also after therapy withdrawal, and rates of JADAS-MDA and JADAS-remission off-biologics were determined.

Safety analysis
Safety was analysed based on AE reporting for all patients during the whole treatment period. AEs and serious AEs (SAEs) were defined according to the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) E6 Section 1.2 [15]. Exposure-adjusted AE rates were calculated per 100 patient-years (PY) with 95% confidence interval (CI). AEs and SAEs were attributed to the etanercept treatment if the patient had been treated with etanercept at the time of the occurrence of the AE or during the last 90 days prior to the AE occurrence, regardless of a possible cotreatment with methotrexate. Malignancies, pregnancies, and deaths were additionally analysed in the ever-treated population.

Statistical analysis
For the comparison of baseline characteristics, the chisquared, Fisher's exact, or Mann-Whitney U test was used, depending on data distribution. Mean changes from baseline in each effectiveness parameter were compared using the unpaired t test. Differences in AE rates were analysed using risk ratios (RRs) and the Wald test. A p value < 0.05 was considered statistically significant. Analyses were conducted with IBM SPSS Statistics version 23 (IBM Corp., Armonk, NY, USA) and SAS 9.3 (SAS Institute, Cary, NC, USA).

Study population
Within the JIA patients who initiated etanercept treatment for the first time between 2001 and 2019, 2725 were eligible for the analysis. Mean etanercept dose was 0.79 ± 0.21/kg/week. The administration regimen was once weekly in 56.9% of patients and twice weekly in 43.1%. Overall, the majority of etanercept-treated patients in Germany showed to be diagnosed with (rheumatoid factor (RF)-negative) polyarthritis, extended oligoarthritis, or ERA, although the cohort subtype distribution varied considerably over the years (  (Table 1). Age at disease onset was comparable between the cohorts, while age at baseline was significantly higher in the etanercept group (12.1 ± 4.4 versus 9.8 ± 4.8 years; p < 0.0001; Table 1). Disease duration at baseline, as calculated from symptom onset to start of cohort defining treatment, was also significantly higher in the etanercept cohort (4.1 ± 3.7

Effectiveness
On etanercept, the mean JADAS10 decreased from 15.3 ± 7.5 at baseline to 5.6 ± 5.7 (p < 0.0001) after 3 months and to 4.1 ± 5.3 (p < 0.0001) after 12 months of treatment (Fig. 2). Patients recruited in the most recent years achieved a lower JADAS10 after 12 months on etanercept compared to those enrolled in the earlier years, although the difference did not reach significance (2.   The effectiveness parameters JADAS-MDA, JADASremission, and ACR-inactive disease were also evaluated in an intention-to-treat (ITT) analysis, with patients who discontinued due to remission considered as responders, and patients discontinuing for any other reason as nonresponders. In this analysis, JADAS-MDA was reached by 40.2, 41.2, and 33.8% of patients at months 3, 12, and 24, and by 21.3, 19.7, and 19.0% after 5, 7, and 9 years. JADAS-remission was reached by 15.1, 25.6, and 23.9% of patients at months 3, 12, and 24, and by 19.7, 19.1, and 18.7% after 5, 7, and 9 years. ACR-inactive disease was reached by 18.8, 27.0, and 25.0% of patients at months 3, 12, and 24, and by 19.6, 19.2, and 18.8% after 5, 7, and 9 years.

Safety
During 5988 patient-years of etanercept exposure, a total of 2053 AEs were reported to the registry (Table 3) Two developed a sacroiliitis, one an iridocyclitis, and one an alpha-1-antitrypsin deficiency. Mean age at arthritis onset was 6.8 ± 4.3 years, and mean age at IBD onset was 13.7 ± 2.7 years. Patients developed IBD after 2.3 ± 1.9 years on ongoing etanercept therapy. All 19 patients had been pretreated with MTX. Rates of psoriasis (0.07 versus 0.003/100PY; p = 0.4) and aggravation/new onset of uveitis (1.9 versus 1.4/ 100PY; p = 0.09) did not differ significantly between the two analysed cohorts. One patient in the etanercept cohort developed demyelination and none in the biologicnaïve control cohort. The lesion, a minor alteration of the periventricular white matter, was discovered incidentally in an asymptomatic patient and has been described earlier [7]. Other reported immune-mediated events were Henoch-Schonlein purpura, leukocytoclastic cutaneous vasculitis, and lupus-like syndrome in one patient each. All three patients were on etanercept treatment.
Fifteen reports of suicidal intention or ideation, suicide attempt, or depression were documented in the etanercept cohort. Six were observed in RF-negative and one in RF-positive polyarthritis patients, two cases in extended and one in persistent oligoarthritis patients, and two in ERA and three in PsA patients. In all, the occurrence of suicide intention/depression was significantly higher in the etanercept group than in the biologic-naïve group (0.25 versus 0.05/100PY; p = 0.04).
Three pregnancies occurred in patients under etanercept treatment at the time of conception. A 17-year-old patient who was treated with etanercept and methotrexate delivered at term a healthy 3360-g male infant after a pregnancy without complications. Treatment was interrupted as her pregnancy was diagnosed at 6 weeks of gestation. The child was developing normally at two months of age. An 18-year-old patient gave birth to a healthy male infant, weight and gestational age of which have not been reported by the documenting physician. At the age of six months, the child showed normal growth and development. The third patient decided on an induced abortion at 12 weeks of gestation. Additionally, a miscarriage after 12 weeks of gestation was reported to the registry in an 18-year-old patient, two years after discontinuation of etanercept. She had been treated with etanercept 50 mg weekly over nine months. The patient received hydroxychloroquine 300 mg daily from 20 months before conception through eight weeks of pregnancy. No pregnancy was recorded in the biologic-naïve cohort.
Three malignancies were documented in patients on etanercept at the time of diagnosis (0.05/100PY). An 18year-old male patient developed a non-familial thyroid carcinoma. One case each of Hodgkin's and non-Hodgkin's lymphoma was reported in two male patients. Malignancies were reported to the registry in five other patients who had been exposed to etanercept in the past: two cases of lymphoproliferative disorder and one case each of anaplastic ependymoma, yolk sac carcinoma, and cervix dysplasia. All patients recovered. In the biologicnaïve cohort with methotrexate, two cases of acute  Data are shown as n (%). Multiple reasons could be given lymphatic leukaemia (ALL) were documented (0.05/ 100PY). One patient recovered; the second died. All malignancy cases have been previously described [16]. In all patients ever treated with etanercept, five deaths were reported, three of these during drug exposure. Two deaths occurred during adolescence and three in adulthood. One patient with systemic JIA died due to septic shock while on treatment with etanercept, after having been pretreated with cyclophosphamide and chlorambucil years before. A second patient with systemic JIA succumbed to heart failure by macrophage activation syndrome (MAS), one year after discontinuation of etanercept due to inefficacy. Both deaths occurred at the age of 16 years and have been formerly reported [17]. Of the three deaths during adulthood, one occurred in a 22-year-old due to perimyocarditis with arrhythmia, eight weeks after voluntary discontinuation of etanercept. A second patient died at the age of 22 years by suicide, seven years after etanercept discontinuation, and a third one died at the age of 23 years due to pseudomembranous enterocolitis by a septic urinary tract infection with renal failure and pancytopenia after 13 years of etanercept exposure. The events were considered as not related to etanercept treatment. In the biologic-naïve group, one death was reported. A 13-year-old female patient on methotrexate succumbed to ALL.

Discussion
The current registry study represents the largest cohort of etanercept-treated JIA patients studied. To our knowledge, it is the first report on safety and effectiveness of etanercept including all JIA categories and following patients up to nine years of continuous treatment. A significant improvement from baseline was observed in all analysed efficacy parameters already after three months of treatment. These improvements were maintained during up to nine years of sustained drug use, in accordance to observations from previous studies [18][19][20]. Patients who had been recruited in most recent years (2016-2018) had higher JADAS-MDA and JADAS-remission responses after one year of treatment compared with patients recruited in earlier years (2001)(2002)(2003). This is likely reflective of the shorter disease duration and lower disease activity at the start of treatment of the formers, supporting the increasing evidence in the literature indicating a positive prognostic effect of an early aggressive treatment due to a suggested window of opportunity [21].
Next to an as-observed analysis evaluating disease activity in patients who continued treatment, we performed an effectiveness analysis in the ITT population. In this analysis, the percentage of patients reaching JADAS-remission or ACR-inactive disease stayed stable over time, while the percentage of patients presenting JADAS-MDA decreased over time. However, this represents a conservative assessment, since it presumes that all patients lost to follow-up did poorly, while only 28.9% of the patients discontinued due to inefficacy or intolerance. Of note, one third of the patients who discontinued due to remission remained in clinical remission off-biologics one year after etanercept withdrawal. Incidence of adverse events was low, and no new safety signal emerged with long-term continuous etanercept exposure. Rates of SAE and serious infection were higher than in biologic-naïve patients, remaining yet low in both groups. By interpreting these data, it should be considered that patients initiating etanercept, according to JIA therapy recommendations, were mostly those who failed to respond or responded inadequately to treatment with ≥ 1 conventional DMAR D, hence, those with a refractory JIA. They presented longer disease duration and higher disease activity, and had received more previous treatments, so one or more of these factors may have accounted for the higher serious infection rate. In fact, as described by Beukelman et al. [22], it is hypothesised that, in JIA patients, the underlying disease process itself, independently from treatment, might increase the risk of serious infections. The comedication with methotrexate, documented in two thirds of the etanercept cohort, may be also a contributing factor, as it was showed in prior studies that serious infections had a higher incidence in patients treated with etanerceptmethotrexate combination treatment compared to monotherapy [3]. The rates for serious infections in the here analysed etanercept cohort were similar to those reported in other long-term registry studies [19,20], but lower than what described in long-term controlled clinical trials [18,23]. While other studies have also found higher serious infection rates upon etanercept [3,24], others have described comparable rates to non-biological treatments. A large study comparing Medicaid data from 2713 new TNFi users found no significant difference in hospitalised infection rates between TNFi and methotrexate treatments [25]. In our study, herpes zoster reactivation was the only opportunistic infection reported, and it was more often observed in etanercept than in biologic-naïve patients with methotrexate, similarly to what described by other groups [23,24].
Incidence of new-onset IBD was higher in etanercept than in biologic-naïve patients, which is consistent with previous findings [26]. Since etanercept was shown to be ineffective in Crohn's disease, gastrointestinal manifestations in patients with IBDassociated arthritis may occur more likely under treatment with etanercept [27]. Rates of other autoimmune disorders, mostly psoriasis and uveitis, were comparable within the two groups. Only one case of suspected demyelination has been reported to our registry in conjunction with etanercept exposure. In the data from the large multinational Pharmachild/ PRINTO pharmacovigilance study, demyelination was also a rare event [28].
Prior studies have suggested that patients with JIA may be at increased risk of malignancy [4]. In our study, treatment with etanercept did not associate with a higher incidence of malignancies. Similar results were reported in a Scandinavian registry study [5] as well as in the Pharmachild registry [28].
The interpretation of the results provided here is potentially influenced by the classic limitations accompanying registry studies, such as lack of blinding, lack of randomisation, and, for the long-term efficacy analysis, lack of an internal comparator. Comparison analyses between the two groups in terms of efficacy were, for the nature itself of the study as a registry analysis and not a randomised controlled clinical trial, not feasible. In addition, the majority of the etanercept patients had already received methotrexate and, as per JIA treatment recommendations, had failed to respond or responded inadequately to it, if a switch or escalation of treatment was required. Moreover, in two thirds of etanercept patients, methotrexate was not discontinued, but etanercept was added to it. Similarly, the biologic-naïve group used as a comparator for the long-term safety assessment presented significant differences at baseline to the etanercept group. This is due to the fact that while the latter had a refractory and more active disease, the former had just begun treatment with methotrexate as firstline DMARD.
Yet, registry studies remain of great importance because they reflect routine care and allow investigation of safety and effectiveness in a complete spectrum of patients and in a real-world setting. Differently, decisions in randomised clinical trials (RCTs) may be influenced by protocol and inclusion/exclusion criteria generate a mostly homogeneous study population of selected patients, e.g. by excluding determinate subtypes, comorbidities, or concomitant drugs. Moreover, in the present analysis, the high number of patients and the long study period allow detection of rare adverse events and adverse events occurring with long-term exposure.

Conclusions
While no increased rates of malignancies and autoimmune disorder other than IBD were observed under etanercept, SAE and serious infection ratios were lower in the biologic-naïve patients with methotrexate, highlighting the high tolerability of the latter. Moreover, long-term etanercept treatment demonstrated a sustained efficacy in this large cohort of JIA patients.