Canakinumab in systemic juvenile idiopathic arthritis: real-world data from a retrospective Italian cohort

Abstract

Objective

The objective of this study was to use real-world data to evaluate the effectiveness and safety of canakinumab in Italian patients with systemic JIA (sJIA).

Methods

A retrospective multicentre study of children with sJIA was performed. Clinical features, laboratory parameters and adverse events were collected at baseline, and 6 and 12 months after starting canakinumab. The primary outcome measure of effectiveness was clinically inactive disease (CID) off glucocorticoids (GCs) treatment at 6 months.

Results

A total of 80 children from 15 Italian centres were analysed. Of the 12 patients who started canakinumab in CID while receiving anakinra, all maintained CID. Of the 68 with active disease at baseline, 57.4% achieved CID off GCs at 6 months and 63.8% at 12 months. In univariate analysis, the variables significantly related to non-response were number of active joints (NAJs) ≥5, history of macrophage activation syndrome (MAS) and disease duration. Multivariate analysis confirmed the association between non-response and NAJs ≥5 [odds ratio (OR) 6.37 (95% CI: 1.69, 24.02), P = 0.006] and between non-response and history of MAS [OR 3.53 (95% CI: 1.06, 11.70), P = 0.039]. No serious adverse events were recorded in this series. There were two cases of MAS during canakinumab, leading to a rate of 2.9 episodes per 100 patient years.

Conclusion

We have confirmed, using real-world data, the efficacy of canakinumab in sJIA in a multicentric cohort. History of MAS and higher NAJ were associated with lower probability of achieving CID.

systemic juvenile idiopathic arthritis, canakinumab, clinically inactive disease

Rheumatology key messages

• Based on real-world data, canakinumab was found to be effective therapy for sJIA patients.

• A history of MAS and a higher NAJs are associated with lower response to canakinumab.

Introduction

Systemic JIA (sJIA) is characterized by chronic arthritis, systemic symptoms, marked elevation of inflammatory markers and absence of autoantibodies [1]. According to ILAR, diagnosis is based on the presence of chronic arthritis associated with daily fever for at least 2 weeks, with intermittent rash, lymphadenopathy, hepatomegaly, splenomegaly or serositis [2]. The prevalence of systemic symptoms and the demonstration of a key role of IL-1 and IL-6 in the pathogenesis of the disease led to the understanding of sJIA as an autoinflammatory rather than an autoimmune disease [3]. Treatment with systemic glucocorticoids (GCs) has been commonly used, but with significant side effects. Because of the side effects, and in view of the important role of the IL-1 and IL-6 pathways in patients with sJIA, IL-1 and IL-6 inhibitors were successfully used in children with sJIA [4–9]. Several studies have confirmed the efficacy of anakinra (a recombinant IL-1 receptor antagonist that blocks both IL1α and IL1β biologic activity) in a significant proportion of patients with sJIA, especially in the first phase of the disease [5, 10–12]. Canakinumab is a fully human, anti-IL1β monoclonal antibody that selectively binds soluble IL-1β, which is overproduced in patients with sJIA [13]. Previous published clinical trials have demonstrated the efficacy of canakinumab in sJIA patients in a substantial number of patients [6, 14–17]. In these studies, a consistent percentage of patients were able to stop GCs treatment, ranging from 15.6 to 36.4% [6, 14–17].

To the best of our knowledge, little real-world data is available regarding the use of canakinumab in patients with sJIA. Data from the German BIKER register for 22 sJIA patients treated with canakinumab showed ACR remission in ∼60% of patients after 24 months of treatment [18]. In that study, patients treated with IL-1 inhibitors (anakinra and canakinumab) within 12 months of the disease commencement achieved clinical remission more frequently than patients who started treatment after the first 12 months [18], in accordance with the hypothesis of window of opportunity [9]. In a retrospective study, 27 out of 168 sJIA patients (16.1%) were treated with canakinumab due to lack of efficacy or local reaction to anakinra, with clinical and laboratory remission in 85% of them [19].

In this report, we describe real-world data with response rates, outcomes and safety in Italian patients with sJIA treated with canakinumab.

Methods

This is a retrospective, observational, multicentre study of 82 patients with sJIA who started canakinumab before the end of 2019.

All centres of the Italian Pediatric Rheumatology Study Group were asked for a census of patients with sJIA treated with canakinumab through completion of data collection forms. Medical records for 82 patients with sJIA from 15 Italian pediatric rheumatology centres were retrospectively reviewed: 76 of the patients had follow-up data for 6 months of canakinumab treatment. Two of the 82 patients were excluded from the analysis: one was lost to follow-up before 6 months, and the other developed lymphopenia so canakinumab was withdrawn. Canakinumab was withdrawn from another 4 patients before 6 months due to lack of efficacy, but they were included in the analysis as non-responders at 6 months.

Clinical and laboratory data were collected in combination with general and demographic data, such as age, gender, age at disease onset, age at starting canakinumab, dosages used, duration of treatment, concomitant and previous treatments, previous macrophage activation syndrome (MAS), and overall adverse events.

Diagnosis of sJIA was established according to ILAR criteria [2]. Data were collected using standardized forms. Demographic data, disease history, and baseline (defined as canakinumab initiation) clinical and laboratory features of the patients are included in Table  1. Presence of fever was defined as body temperature above 38°C. The primary outcome was clinically active disease (CID) at 6 months, according to the preliminary criteria for inactive disease and clinical remission of JIA [20] without glucocorticoids (GCs) treatment.

Table 1

Demographic data and baseline characteristics of patients

	All patients (N = 80)	IL-1 inhibitor naïve (N = 35)	Previous IL-1 inhibitor with AD (N = 33)	Previous IL-1 inhibitor in CID (N = 12)
Gender, femalea	49 (61.3)	22 (62.9)	20 (60.6)	7 (58.3)
Age at disease onset (yrs)b	6.5 (3.8–10.6)	7.9 (4.1–11.3)	5.7 (3.0–9.5)	7.1 (4.4–11.4)
Age at treatment start (yrs)b	10.4 (6.2–13.4)	10.3 (6.2–12.4)	10.1 (6.1–15.3)	12.3 (9.9–15.3)
Previous treatment
Glucocorticoidsa	74 (92.5)	32 (91.4)	32 (97.0)	10 (83.3)
DMARDs
MTXa	26 (32.5)	13 (37.1)	11 (33.3)	2 (16.7)
Ciclosporina	19 (23.8)	7 (20.0)	8 (24.2)	4 (33.3)
bDMARDsa	50 (62.5)	5 (14.3)c	33 (100)d	12 (100)e
Previous MASa	26 (32.5)	10 (28.6)	9 (27.3)	7 (58.3)
Time from onset to receiving canakinumab (mo)b	20.7 (5.9–48.0)	8.3 (2.8–31.7)	26.3 (10.9–94.5)	31.4 (12.4–83.7)
Baseline features
Fever (>38.0°C)a	38 (47.5)	19 (54.3)	19 (57.6)	0 (0.0)
Rasha	26 (32.5)	14 (40.0)	12 (36.4)	0 (0.0)
Number of active jointsb	0 (0–4)	2 (0–5)	1 (0–4)	0 (0–0)
CRP (mg/dl)b	2.39 (0.33–9.00)	5.45 (0.65–10.83)	2.37 (0.60–13.10)	0.18 (0.07–0.31)
Ferritin (ng/ml)b	137 (62–480)	200 (120–909)	165 (73–358)	34 (31–44)
Neutrophil count (×10³/µl)2	6.5 (3.5–10.0)	7.2 (5.0–10.4)	5.8 (3.5–8.0)	3.5 (2.6–12.5)
Haemoglobin (g/dl)b	11.6 (10.6–12.8)	11.1 (10.4–12.0)	11.4 (10.6–12.5)	13.0 (12.3–13.5)
Platelet count (×10³/µl)b	343 (260–473)	397 (280–499)	324 (271–428)	245 (213–331)
Concomitant treatment at baseline:
Glucocorticoidsa	54 (67.5)	30 (85.7)	24 (72.7)	0 (0.0)
Glucocorticoid dose (mg/kg/day)b	0.5 (0.3–1.0)	0.5 (0.4–1.0)	0.5 (0.3–0.9)	–
DMARDs
MTXa	14 (17.5)	10 (28.6)	3 (9.1)	1 (8.3)
Ciclosporina	10 (12.5)	5 (14.3)	2 (6.1)	3 (25.0)
Canakinumab dose (mg/kg/4 w)b	4.0 (3.3–4.0)	4.0 (2.8–4.0)	3.9 (3.7–4.0)	4.0 (3.1–4.0)

Data for all 80 patients, divided into groups based on IL-1 inhibitor therapy (naïve or previous therapy) and the state of the disease at the start of canakinumab. ^aNumber (percentage); ^bMedian (first and third quartiles); ^c1 etanercept, 4 tocilizumab; ^d25 anakinra, 2 etanercept + anakinra, 2 tocilizumab + anakinra, 1 anakinra + tocilizumab + adalimumab, 1 etanercept + infliximab + tocilizumab + anakinra, 1 infliximab + golimumab + tocilizumab + anakinra, 1 abatacept + etanercept + golimumab + adalimumab + tocilizumab + rituximab + anakinra; ^e10 anakinra, 1 anakinra + tocilizumab; 1 etanercept + tocilizumab + anakinra. AD: active disease; CID: clinically inactive disease; bDMARDs: biologic DMARDs; MAS: macrophage activation syndrome.

The normal ranges used for laboratory data were as follows: CRP <0.5 mg/dl, ESR <15 mm/h, ferritin <450 ng/ml, white blood cell count 5500–15 000/µl, neutrophil cell count 1650–8250/µl, haemoglobin 10.5–15.5 g/dl and platelet count 150–450 × 10³/µl. The study was approved by the Ethics Committee of the IRCCS Ospedale Pediatrico Bambino Gesù (1683 OPBG 2018). Patients and/or their parents provided written informed consent.

Statistical analysis

Qualitative variables were expressed as absolute frequency and percentage. Proportions were compared by the χ² test or Fisher’s exact test, as appropriate. Quantitative variables, reported as medians and interquartile ranges (first and third quartiles), were analysed using the Mann–Whitney U test for unmatched groups. Association of baseline and disease history variables with the primary outcome (i.e. response defined as CID off GCs) was assessed with univariate analysis. Variables with P < 0.05 in univariate analysis were included in a logistic regression model having as a dependent variable non-response after 6 months of therapy with canakinumab.

All statistical tests were two sided; a P-value of <0.05 was considered to be statistically significant. The analyses were performed and graphs generated using Stata 15.1 software (StataCorp LLC, College Station, Texas USA, 2017) and GraphPad Prism 5.0 (GraphPad Software, Inc., San Diego, CA).

Results

Patients

The demographic and laboratory features of the 80 patients analysed (76 patients with a follow-up of 6 months, and 4 patients for whom canakinumab treatment was withdrawn before 6 months due to lack of efficacy) are summarized in Table  1. At baseline, 35 patients (43.7%) were IL-1 inhibitor–naïve and 45 (56.3%) had been previously treated with the IL-1 inhibitor anakinra. Among the patients previously treated with anakinra, 33 (73.3%) discontinued anakinra due to lack of efficacy, and they had active disease (AD) when switched to canakinumab treatment. Twelve patients (26.7%) on anakinra had CID when switched to canakinumab due to discomfort and/or pain from daily s.c. injections (Table  1). The median age at disease onset was comparable among the three groups. Almost all patients had previously received GCs, while 47.5% of patients had received one or more DMARDs (32.5% MTX and 23.8% CSA) (Table  1).

Of the 68 patients who started canakinumab during AD, 38 (55.9%) had fever. The median number of active joints (NAJs) was 2. It is worth noting that among the patients who started canakinumab after having failed to respond to anakinra, 4 patients had failed multiple lines of biologics with varying mechanisms of action (see footnote in Table  1). The median dose of canakinumab was 4.0 mg/kg every 4 weeks.

Response at 6 months

After 6 months of canakinumab treatment, 51 patients (63.7%) met the criteria for CID off GCs, while 29 patients (36.3%) had AD (14 previously treated with anakinra and 15 biologic naïve) (Fig.  1). All 12 patients treated with anakinra who switched to canakinumab, while being in CID off GCs, maintained CID off GCs at 6 months. Of the 33 patients previously treated with anakinra who switched to canakinumab while having AD, 19 (57.6%) patients achieved CID at 6 months, a rate comparable with that of the 20 IL-1 inhibitor–naïve patients (57.1%) (Fig.  1).

Fig. 1

Patients’ disposition and achievement of clinically inactive disease (CID) off GCs

Univariate analysis of features associated with response

In order to evaluate whether the 6-months response to canakinumab therapy was related to baseline features, we excluded from the analysis the 12 patients who started canakinumab during CID while receiving anakinra. The 68 remaining patients were categorized as responders (39/68, 57.4%) or non-responders (29/68, 42.6%), and the variables were evaluated in univariate analysis. The four patients who stopped canakinumab before 6 months from baseline due to its lack of efficacy were considered as non-responders by the treating physicians and therefore as non-responders at 6 months in the analysis. As shown in Table  2, no difference in outcome was found to be related to the presence of fever at baseline. Notably, a baseline variable that was significantly related to the response to canakinumab was the NAJs, which was higher in the non-responders (P = 0.016) (Table  2). Based on the accepted clinical definition of polyarticular JIA, we chose a cut-off of 5 active joints. With this cut-off, we found a statistically significant difference (P = 0.001) between non-responders (44.8%) and responders (10.3%) (Table  2).

Table 2

Univariate analysis for predictors of CID off glucocorticoids at 6 months in patients with AD at baseline

	Responders (N = 39)	Non-responders (N = 29)	P-value
Gender, femalea	21 (53.9)	21 (72.4)	0.12b
Age at disease onset (yrs)d	6.1 (3.9–11.1)	6.6 (3.7–9.9)	0.63
Age at treatment start (yrs)d	9.9 (6.0–13.3)	10.4 (8.4–15.3)	0.31
Previous treatment with anakinraa	19 (48.7)	14 (48.3)	0.97b
History of MASa	7 (18.0)	13 (44.8)	0.016b
Time from onset to receiving canakinumab (mo)d	13.9 (3.0–33.4)	28.7 (6.9–89.7)	0.047
Baseline features
Fever (>38.0°C)a	22 (56.4)	16 (55.2)	0.92b
Rasha	16 (41.0)	10 (34.5)	0.58b
Number of active jointsd	1 (0–2)	4 (0–7)	0.016
Active joints ≥5a	4 (10.3)	13 (44.8)	0.001b
CRP (mg/dl)d	2.70 (0.60–10.83)	5.45 (0.80–13.45)	0.40
Ferritin (ng/ml)d	190 (125–480)	133 (77–924)	0.71
Neutrophils count (×10³/µl)d	6.2 (4.5–9.3)	7.4 (4.1–12.5)	0.67
Haemoglobin (g/dl)d	11.4 (10.5–12.6)	11.1 (10.5–12.2)	0.49
Platelets count (×10³/µl)d	355 (280–489)	356 (261–444)	0.57
Concomitant treatment at baseline
Glucocorticoidsa	28 (71.8)	26 (89.7)	0.07b
Dose of GCs (mg/kg/day)d	0.5 (0.4–0.9)	0.6 (0.3–1.0)	0.86
DMARDs
MTXa	5 (12.8)	8 (27.6)	0.13b
Ciclosporina	1 (2.6)	6 (20.7)	0.037c
Canakinumab dose (mg/kg/4w)d	3.9 (3.5–4.0)	3.9 (3.0–4.0)	0.50

Univariate analysis for predictors of clinically inactive disease (CID) off glucocorticoids (GCs) therapy at 6 months from baseline in patients who started canakinumab with active disease (AD) (n = 68).^aNumber (percentage): ^cχ² test; ^dFisher’s exact test. ^bMedian (first and third quartiles)—Mann–Whitney U test. MAS: Macrophage activation syndrome.

The risk of non-response after 6 months of therapy with canakinumab among the 17 patients with NAJs ≥5 at baseline was 2.4 times higher than the risk of non-response among the 51 patients with NAJs <5 at baseline [risk ratio 2.4 (95% CI: 1.5, 4.0); P = 0.001] (Fig.  2A). Responder patients started canakinumab earlier in their disease course than non-responders (P = 0.047). History of MAS was significantly more frequent in non-responders (44.8%) than in responders (18.0%) (P = 0.016). Patients with a history of MAS presented a risk of non-response 2.0 times higher (95% CI: 1.2, 3.3) than patients without history of MAS (Fig.  2B). All but one of the patients receiving ciclosporin at baseline had a history of MAS, and, as expected, ciclosporin treatment was more frequent in non-responders.

Fig. 2

Response after 6 months of therapy: relationship with NAJ at baseline (A) and history of MAS (B)

(A) Patients with NAJ <5 at baseline had a significantly better response (68.6% vs 23.5%; chi-square test: P = 0.001). Risk ratio of non-response at 6 months of therapy with canakinumab based on NAJ ≥5 was 2.4 (95% CI: 1.5, 4.0). (B) Patients without a history of MAS had a significantly better response (66.7% vs 35.0%; chi-square test: P = 0.016). Risk ratio of non-response at 6 months of therapy with canakinumab based on history of MAS was 2.0 (95% CI: 1.2, 3.3). MAS: macrophage activation syndrome; NAJ: number of active joints.

Logistic analysis of features associated with response

A logistic regression analysis was performed including NAJs ≥5, history of MAS and disease duration from onset to baseline. Treatment with ciclosporin at baseline was excluded because of collinearity with history of MAS. The full model (pseudo R²: 0.18) showed that NAJs ≥5 [OR 6.37 (95% CI: 1.69, 24.02), P = 0.006) and history of MAS [OR 3.53 (95% CI: 1.06, 11.70), P = 0.039] remained statistically significant, confirming their association with non-response, independently of disease duration [OR 1.00 (95% CI: 0.99, 1.01), P = 0.29] (Table  3). The predicted margins (average predicted response) were calculated to assess the probability of response after 6 months of therapy with canakinumab, considering the combination of the two statistically significant variables identified in the logistic model. Patients with a history of MAS and NAJs ≥5 had an 11.8% (95% CI: 0.0, 26.9) probability of response, while patients without a history of MAS and with NAJs <5 had a 75.9% (95% CI: 62.7, 89.0) probability of response (Fig.  3).

Fig. 3

Logistic regression analysis of non-achievement of clinically inactive disease at 6 months from starting canakinumab

Best-fitting model obtained through logistic regression procedures (pseudo R²: 0.18). The dependent variable was non-achievement of clinically inactive disease at 6 months from canakinumab start. MAS: macrophage activation syndrome.

Table 3

Best-fitting model obtained through logistic regression procedures (Pseudo R²: 0.18)

Explanatory variable	OR (95% CI)	Pa
Number of active joints ≥5	6.37 (1.69, 24.02)	0.006
History of MAS	3.53 (1.06, 11.70)	0.039
Disease duration from onset to baseline	1.00 (0.99, 1.01)	0.29

The dependent variable was non-achievement of clinically inactive disease at 6 months from canakinumab start. ^aBy likelihood ratio test. MAS: macrophage activation syndrome.

Response at 12 months

Fifty-seven patients reached a follow-up of at least 12 months (Fig.  1). Of the 36 patients in CID at 6 months and with a follow-up visit at 12 months, 32 (88.9%) maintained CID at 12 months and only 4 (11.1%) flared. Of the 29 patients with AD at 6 months, 21 reached a follow-up of 12 months; 6 (28.6%) of them achieved CID at 12 months. Overall, at 12 months 38/57 patients (66.7%) were in CID off GCs, and 19/57 (33.3%) had AD. When we evaluated the predictive factors identified at 6 months, at 12 months the frequency of NAJs ≥ 5 was significantly (P = 0.041) higher in non-responders (47.1%) than in responders (16.7%). At 12 months, the difference in history of MAS (responders 16.7%; non-responders 35.3%) showed the same trend, but the difference did not achieve statistical significance (P = 0.17), possibly due to the smaller sample size.

Safety

No serious adverse events were recorded in this retrospective series. One patient developed lymphopenia, which led to withdrawal of canakinumab after 3 months. We observed one patient with a mild injection site reaction, 6 patients with infections, none of them reported as serious, one patient with herpes zoster and one case of transient neutropenia. All resolved with standard treatment and did not require discontinuation of canakinumab. Among the 80 patients, there were two cases of MAS during treatment with canakinumab, leading to a rate of 2.9 MAS episodes per 100 patient-years. Both of these patients developed MAS within 12 months of starting canakinumab, and one of them had a previous episode of MAS before starting the therapy.

Discussion

There are no real-world reports on a large series of patients with sJIA treated with canakinumab outside of the setting of controlled trials [6, 14, 15]. In our real-world setting, we chose to use as the primary efficacy outcome the achievement of CID off GCs after 6 months of therapy with canakinumab, according to the criteria proposed by Wallace et al. [20]. This primary outcome was used to identify patients with a clinically relevant response that combined absence of disease symptoms with withdrawal of glucocorticoids. Indeed, high rates of responses have been demonstrated in the trials with canakinumab, as well as with other IL-1 inhibitors in sJIA [10–12, 21]. In our study, we found that canakinumab treatment led to CID in ∼57% of sJIA patients at 6 months and in 67% of those who achieved a 12 months follow-up. This is consistent with what has been reported in a smaller number of patients in a real-world multicentric setting from the BIKER register, in which ∼60% (13/22) of sJIA patients achieved persistent CID (ACR remission) after 24 months of treatment [18].

The percentage of patients with CID off GCs found in our study is higher than that reported in the clinical trials of Ruperto et al. [6, 14, 15], of Nishimura et al. [16] and of Brunner et al. [17]. A small phase II multicentre open-label study [14] showed that 15 days after receiving a dose of canakinumab, 4 of 25 patients (16.0%) achieved CID. A double-blind trial with 177 sJIA patients showed that, after a median of four injections of canakinumab, 31% of patients met CID criteria [6]. A long-term follow-up (up-to 5 years) of 144/177 patients enrolled in these trials showed that 33% of patients achieved CID at 6 months and 40% at 2 years [15]. The open-label phase III trial with 19 Japanese patients showed that 75% of them achieved CID and 31.3% achieved CID off GCs at 12 months [16]. A prospective open-label study of 123 sJIA patients showed that 22% of them met CID (according to the ACR criteria) within 15 days of starting canakinumab, and 31% of them met CID at 6 months [17]. This apparent discrepancy, with a lower rate of CID in clinical trials, may likely be explained by different disease severity at baseline, e.g. the baseline NAJs was 10.0 in the trials but was 2.0 in our cohort.

Furthermore, in our study we showed that canakinumab maintained CID in those patients who had previously achieved CID while receiving anakinra. The patients previously treated with anakinra with AD at baseline achieved CID off GCs at 6 months (19/33, 57.6%) with a rate very similar to those who were IL-1 inhibitor naïve (20/35, 57.1%). This observation suggests that failure of one anti-IL1 drug does not necessarily preclude use of another one. The mechanistic implications of our observation remain unclear. Although both anakinra and canakinumab are IL-1 inhibitors, their mechanisms of action are not superimposable: anakinra blocks both IL-1α and IL-1β, while canakinumab binds only to IL-1β. Moreover, these drugs have different pharmacokinetic characteristics. Canakinumab, as expected for monoclonal antibodies targeting soluble molecules, shows linear, dose-proportional pharmacokinetics. The effect of anakinra, which is a competitive receptor antagonist, depends on the dose of drug, on the concentration of IL-1, on its competitiveness and on the level of spare receptors [22]. Whether these different features translate into different efficacy in different patients remains unknown.

Within the limitation of the relatively small number of patients, we have analysed the relationship of the clinical variables to response. We found a trend towards early treatment being associated with better response to canakinumab. This is in line with the findings of the BIKER registry, in which patients who started IL-1 inhibitors within 12 months of disease onset met JADAS remission more frequently than patients who started treatment later [18]. This is also consistent with other experiences with anakinra [12, 23]. Even if formal demonstration of a window of opportunity in a clinical trial is not available, our observation adds to the increasing body of evidence, albeit uncontrolled, that early treatment is associated with better response.

When predictors of lower probability of response were analysed, we found that NAJs ≥5 and history of MAS were associated with non-response in univariate and multivariate analysis. This is consistent with some observations [23, 24] (but not all) using anakinra in sJIA, although with a different cut-off [12]. In our study, patients with NAJs ≥5 at baseline had a significantly higher risk of non-response. Regarding history of MAS, an association with non-response to IL-1 inhibitors has not previously been reported. However, Hinze et al. recently reported that C-X-C Motif Chemokine Ligand 9 (CXCL9) levels were higher in sJIA patients who did not achieve complete response with canakinumab [25]. CXCL9 is directly related to IFN-gamma (IFN-γ). Mounting evidence points to a pathogenic role of IFN-γ in MAS, and other hemophagocytic lymphohistiocytosis (HLH) forms [26–28]. In patients with a history of MAS, we found higher levels of CXCL9 [26]. Altogether, these observations suggest that the biology of patients at risk of MAS may confer a lower probability for complete response to canakinumab. Further studies on the immunobiology of sJIA and MAS are needed. The data obtained with a predicted margin analysis showed that patients with a history of MAS and NAJs ≥5 had a low probability of response with a 95% CI: 0, 26.9%. This appears to be of potential clinical relevance in guiding therapeutic decisions, although this relates to a small percentage of patients (in our series there were 7 out of 68, all non-responders). This result needs to be confirmed in a different and possibly larger series of patients with sJIA.

Regarding the safety of canakinumab, no serious adverse events were recorded. The adverse events reported were mild injection site reactions, non-serious infections and one case of herpes zoster. None of these events required discontinuation of treatment. There was one case of lymphopenia that led to discontinuation of canakinumab before 6 months of treatment. In addition, we observed one case of neutropenia not associated with infection. No safety alerts were detected other than those previously reported in clinical trials and other real-world settings of patients with sJIA treated with canakinumab [6, 14–17, 29]. There were two cases of MAS, leading to a 2.9 rate per 100 patient-years of MAS, similar to the rate of 2.8 reported in the canakinumab clinical trials [30].

In addition to the relatively small number of patients, the limitations of our study include its retrospective nature, with no standardization of other treatments. However, the study reflects the variety of therapies and approaches in a number of different centres, making the results relevant to routine clinical practice.

In conclusion, we confirm in a real-world setting that canakinumab is an effective and safe drug in patients with sJIA. The rate of CID of ∼60% in this study appears to be higher than that of clinical trials in which patients with very severe disease were enrolled.

Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this article.

Disclosure statement: F.D.B. has received research support paid to his institution from AbbVie, Hoffmann-La Roche, Pfizer, NovImmune, Novartis, Sobi and Sanofi. A.R. has received grant support and/or speaking or consultant fees from Angelini, AbbVie, Bristol-Myers Squibb, Johnson & Johnson, Novartis, Pfizer, Reckitt-Benkiser, and Roche. The other authors have declared no conflicts of interest.

Data availability statement

The data underlying this article cannot be shared for the privacy of the individuals that participated in the study. The data will be shared on reasonable request to the corresponding author.