In this retrospective observational study patients were included if they: i) had JIA according to the International League of Associations for Rheumatology (ILAR) criteria [7]; ii) were seen at the IRCCS Istituto Giannina Gaslini of Genoa, Italy, within 6 months after disease onset, defined as the time of occurrence of the first symptoms consistent with JIA, between January 2007 and December 2017; and iii) were followed for a minimum of 6 months after baseline visit. Data were also collected, whenever available, at subsequent visits after 12, 24 and 60 months from baseline. For sake of simplicity, patients were grouped into the following four “functional phenotypes”: systemic arthritis (including patients with systemic arthritis), oligoarthritis (including patients with persistent oligoarthritis), polyarthritis (including patients with extended oligoarthritis and rheumatoid factor, RF-positive and RF-negative polyarthritis), and other arthritis (including patients with enthesitis-related arthritis, psoriatic arthritis, and undifferentiated arthritis).

The study protocol was approved by the Ethics Committee of Regione Liguria (Genoa, Italy) procedure number 642/2022—DB id 12,828, dated 16 June 2023.

During the study period, a step-up approach was adopted in most patients. Patients with oligoarthritis were initially treated with intra-articular glucocorticoids (IAGCs) in all affected joints, without a conventional synthetic disease-modifying antirheumatic drug (csDMARD) in case of involvement of one or two large joints, especially the knees, or together with a csDMARD in case of arthritis in three or four joints or involvement of ankle or wrist joints. Treatment of patients with polyarthritis was also usually started by administering IAGCs in all affected joints, always in association with a csDMARD. The exception was represented by patients with diffuse symmetric polyarthritis, especially if RF-positive, or with involvement of cervical spine or hip joints, who were often given a bridging therapy with systemic glucocorticoids (GCs), always in combination with a csDMARD.

If inactive disease was not reached within 3 to 6 months, treatment was escalated by introducing a biologic DMARD (bDMARD), usually a TNF inhibitor, except for patients with oligoarthritis treated only with IAGC, who were first given a csDMARD and, in case of persistent nonresponse, a bDMARD. Patients with ERA and psoriatic arthritis were treated with a similar approach, depending on the severity and extent of joint disease.

IAGC injections were often repeated after a minimum of 4–6 months in patients with arthritis flares. Methotrexate was the preferred csDMARD for oligoarthritis and polyarthritis, while sulfasalazine was favored for enthesitis-related arthritis.Patients with systemic arthritis were given systemic GCs initially, with quick addition of a bDMARD, generally an IL-1 inhibitor, in case of inadequate improvement of flare during tapering or after discontinuation of systemic GCs. However, in the more recent years we started IL-1 inhibition upfront in many patients with systemic JIA, especially those with more prominent extra-articular features and few or no affected joints. The second-line bDMARD was an alternative IL-1 inhibitor or an IL-6 blocker.

The following baseline information was obtained by reviewing clinical charts: sex, age at disease onset and at first visit, ILAR category, and disease duration. Patients were defined as being antinuclear antibody (ANA)-positive if they had at least 2 positive determinations made at least 3 months apart during follow-up, based on indirect immunofluorescence on Hep-2 cells at a titer of ≥ 1:160. Data extracted at each study visit included presence of active systemic manifestations (fever, skin rash, splenomegaly, generalized lymphadenopathy, serositis) and active uveitis (based on the judgement of the ophthalmologist who performed the evaluation), physician’s global assessment of overall disease activity (PhGA) on a 21-numbered circle visual analog scale (VAS, where 0 = no activity and 10 = maximum activity) [8], and active joint count (AJC), assessed in 73 joints [9]. A joint was defined as active if it displayed swelling or, in the absence of clinically detectable swelling, pain on motion/tenderness and limited range of motion. Laboratory indicators of inflammation included erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP). The medications received by the patients between each study visit were recorded. Data were extracted by five pediatric rheumatology trainees (AIRG, SO, FR, EA and VN) under the supervision of the senior investigator (AR).

The state of CID was defined according to the 2004 Wallace criteria [10], as no joint with active arthritis, no systemic manifestations attributable to JIA, no active uveitis, normal acute-phase reactants, and PhGA indicating no disease activity (defined as score of 0 on the 0–10 VAS). However, for a number of patients the full set of Wallace criteria could not be applied due to the lack of the PhGA. For visits where this parameter was not available, but all the other Wallace criteria were met, the absence of disease activity was inferred, as done in previous studies [11, 12], through the review of the patient chart by consensus of two investigators (AIRG and VN). To substantiate this judgement, the caring physician who originally examined the patient at the time of the visit was asked to review independently his/her clinical notes and confirm the state of inactive disease. Any disagreement between the investigators and the caring physician was resolved by consensus with a senior author (AR).

Descriptive statistics were reported as medians with first and third quartiles (1st–3rd q) for quantitative variables or as absolute frequencies and percentages for categorical variables. Comparison of continuous variables between two groups of patients was made by means of the Mann–Whitney U test. Categorical data were compared by means of the chi-square test or Fisher’s exact test in case of expected frequencies < 5. The Bonferroni adjustment was applied as a correction for multiple comparisons to explore post hoc differences between pairs of patient groups. As the aim was not to compare the frequency of CID and therapeutic interventions across functional phenotypes and because the differences could be easily captured visually, these figures were interpreted only qualitatively.

The software Stata 11 (Stata, College Station, Texas, USA) was used for all statistical analyses.