Twenty-seven patients were included in this study: 21 from Bordeaux teaching hospital, 4 from Dax hospital, and 2 from Pau hospital. Patient characteristics at diagnosis are described in Table 1.
All patients fulfilled both the Yamaguchi and Fautrel criteria. No patients had any personal or familial history of rheumatic, autoimmune, inflammatory, or neoplastic disease. All patients presented with a fever above 38 °C, mainly intermittent and vesperal fevers. The next most prevalent symptoms were arthralgia (96%), arthritis (52%), skin rash (81%), general deterioration (72%), and ear-nose-throat (ENT) symptoms (69%).
Eleven (40.7%) patients were classified as systemic form of disease, and 16 (59.3%) as chronic articular form.
All patients presented with elevated CRP levels. The next most commonly observed anomalies were neutrophilia (96%), low albumin (75%), and increased ferritin (73%). No patients presented with increased creatine kinase (CK), positive extractable nuclear antigen (ENA), positive antineutrophil cytoplasmic antibody (ANCA), positive anticitrullinated protein antibody (ACPA), or positive anticardiolipin. Nine patients were either ANA-positive (five, 1/160), anti-DNA positive (one), or rheumatoid factor (RF)-positive (three) at diagnosis, but these results were judged clinically insignificant because lower than two times the upper normal range and never associated with symptoms concordant with SLE, APLS, or RA and were all controlled to be negative. Furthermore, all patients included in the study met both Yamagushi and Fautrel criteria.
The median duration of disease was 5 years (IQR 2–7) and 6.5 years (IQR 5–9) for the systemic and chronic articular forms, respectively.
All patients presented a chronic refractory form of the disease, which required the off-label use of bDMARD because of failures of corticosteroids and csDMARD or necessity to reduce or stop corticosteroids.
Biological treatment efficacy
Table 2 presents the detailed responses to anakinra and tocilizumab among the patients included in the study. Twenty-six of 27 patients (96.3%) achieved remission with either anakinra or tocilizumab. The median delay in the introduction of the first biotherapy was 5 months (interquartile range (IQR), 1–21 months): this delay was median 1 month (IQR, 0–6) in the systemic form and median 17 months (IQR, 4–38) in the chronic articular form.
A total of 15 patients received anakinra; 5 as a first-line treatment, 4 as a second-line treatment, 5 as a third-line treatment, and 1 as a fourth-line treatment. The median delay between diagnosis and the introduction of anakinra was 1.5 months (IQR, 0–14). Thirteen (86.7%) patients were responders. The 2 anakinra nonresponders received anakinra as a third-line treatment: one had previously failed to respond to tocilizumab, and the other later responded to tocilizumab and even managed to stop it.
A total of 17 patients received tocilizumab, 2 as a first-line treatment, 6 as a second-line treatment, 5 as a third-line treatment, and 4 as a fourth-line treatment. The median delay between diagnosis and the introduction of tocilizumab was 17 months (IQR, 4–39 months). Fourteen patients (82.4%) were responders. Among the 3 tocilizumab nonresponders, one presented with a systemic form and received tocilizumab as a first-line treatment but later responded to anakinra, one presented with a systemic form and received tocilizumab as a fourth-line treatment after responding to anakinra but was switched because of a suspicion of anakinra-induced toxidermia, and one presented with a chronic articular form and received tocilizumab as a fourth-line treatment after failure of anakinra.
The difference of median duration of disease at the introduction of the first bDMARD in the anakinra-treated patients group and in the tocilizumab-treated patients group was statistically significant (p = 0.4457): median 2 months (IQR, 0.25–12 months) in the anakinra-treated patients and median 14.5 (IQR, 3.25–38.25 months) in the tocilizumab-treated patients.
Eight patients (29.6%) managed to stop their biological treatment without relapse at the last known visit which ranged from 6 to more than 24 months after biological treatment discontinuation: 3 of 13 (23%) anakinra responders and 5 of 14 (35.7%) tocilizumab responders. The difference in delays in initiation of anakinra or tocilizumab was not statistically significant between those who managed to stop their biotherapy without relapse versus those who could not stop their biotherapy or those who relapsed, with a median delay of 3 months (IQR, 1.25–13.75) versus 8.5 months (IQR, 1–23.25), respectively (p = 0.61).
Five patients received both biotherapies. Two of these patients presented with a systemic form and were tocilizumab nonresponders but anakinra responders, one presented with a chronic articular form and was anakinra nonresponder but tocilizumab responder, one presented with a chronic articular form and did not respond to either biotherapy, and one patient presented with a systemic form and responded to both biotherapies; this patient first responded to anakinra but corticosteroids could not be tapered so he was switched to tocilizumab in order to successfully stop corticosteroids.
One patient received canakinumab and failed to respond but was anakinra responder. Three patients received antitumor necrosis factor (TNF) agents but these treatments were not efficacious (2 etanercept, 1 infliximab): two of them presented a chronic articular form and one a systemic form.
One of 27 (3.7%) patients did not respond to either biotherapy (anakinra then tocilizumab) and died of a probably infectious-induced AOSD flare.
Predictive factors for preferential responses to anakinra or tocilizumab
Patients who responded to either biotherapies or no biotherapy were excluded from this analysis (n = 2). Table 3 describes the univariate analysis for potential predictive factors of a good therapeutic response to anakinra or tocilizumab. The outcome of interest was the full effectiveness of the treatment, defined as the resolution of all initial systemic and articular symptoms and biological anomalies.
A chronic articular phenotype and the presence of arthritis were associated with a substantial response to tocilizumab: 12 of the 14 patients presenting a chronic articular phenotype responded to tocilizumab whereas 3 of the 13 patients presenting a systemic phenotype responded to tocilizumab (p = 0.0009, OR 36 [2.6–1703]) and 10 of the 14 patients presenting arthritis responded to tocilizumab whereas 3 of the 13 patients without arthritis responded to tocilizumab (p = 0.017, OR 10 [1.22–92.6]).
A systemic form and the absence of arthritis were associated with a substantial response to anakinra: 1 of the 14 patients presenting a chronic articular phenotype responded to anakinra whereas 9 of the 13 patients presenting a systemic phenotype responded to anakinra (p = 0.0009, OR 36 [2.6–1703]) and 3 of the 14 patients presenting arthritis responded to anakinra whereas 9 of the 13 patients without arthritis responded to anakinra (p = 0.017, OR 10 [1.22–92.6]).
No difference in biological variables was statistically predictive of treatment response, although the presence of thrombocytosis seemed to be preferentially associated with a better response to tocilizumab than to anakinra (n = 8/14 vs. n = 3/13, p = 0.123).
Likelihood of steroid withdrawal depending on the specific biotherapy and delays in initiation of biotherapy
All patients received corticosteroids at diagnosis. Among the 26 patients in remission, 14 (53.8%) achieved corticosteroid withdrawal within a mean delay of 27 months, 10 (71.4%) were on tocilizumab, and 4 (28.6%) were on anakinra (Table 2).
Table 4 presents the likelihood of corticosteroid withdrawal depending on treatment with anakinra and tocilizumab: tocilizumab was associated with an increased likelihood of corticosteroid withdrawal (p = 0.029), whereas anakinra did not seem to increase the likelihood of corticosteroid withdrawal (p = 1.000).
A shorter delay before the initiation of anakinra or tocilizumab was not associated with a greater likelihood of corticosteroid withdrawal among biotherapy responders. Among anakinra responders, the difference in the median delay from diagnosis to anakinra initiation between those who achieved steroid withdrawal (n = 4) and those who did not (n = 10) was not statistically significant (p = 0.5767): the median delay was 2.5 months (IQR, 0–8 months) among those who did not stop corticosteroids versus 2 months (IQR, 2–14 months) among those who stopped corticosteroids. Among tocilizumab responders, the difference in the median delay from diagnosis to tocilizumab initiation between those who achieved steroid withdrawal (n = 11) and those who did not (n = 5) was not statistically significant (p = 0.9548): the median delay was 17 months (IQR, 9–21 months) among the tocilizumab responders who did not stop corticosteroids versus 17 months (IQR, 4–38 months) among those who stopped corticosteroids. Initiating tocilizumab as the first-line treatment rather than as a second-line or later treatment did not improve the likelihood of corticosteroid withdrawal (n = 1/2 vs. n= 11/15, p = 1.000). In addition, initiating tocilizumab as a first- or second-line treatment rather than as a third-line or later treatment also did not improve the likelihood of steroid withdrawal (n = 6/8 vs. n = 5/9, p = 0.619); similarly, initiating tocilizumab as a first-, second- or third-line treatment did not improve the likelihood of steroid withdrawal compared to that from tocilizumab used as a fourth-line treatment (n = 9/13 vs. n = 2/4, p = 0.584).