Systemic juvenile idiopathic arthritis and adult-onset Still’s disease are the same disease: evidence from systematic reviews and meta-analyses informing the 2023 EULAR/PReS recommendations for the diagnosis and management of Still’s disease

Abstract

Objectives

To analyse the similarity in clinical manifestations and laboratory findings between systemic juvenile idiopathic arthritis (sJIA) and adult-onset Still’s disease (AOSD).

Methods

Three systematic reviews (SR) were performed. One included cohort studies comparing sJIA versus AOSD that described clinical and biological manifestations with at least 20 patients in each group (SR1). The second identified studies of biomarkers in both diseases and their diagnostic performance (SR2). The last focused on diagnostic biomarkers for macrophage activation syndrome (MAS, SR3). Medline (PubMed), Embase and Cochrane Library were systematically searched. The risk of bias was assessed with an adapted form of the Hoy scale for prevalence studies in SR1 and the Quality Assessment of Diagnostic Accuracy Studies-2 in SR2 and SR3. We performed meta-analyses of proportions for the qualitative descriptors.

Results

Eight studies were included in SR1 (n=1010 participants), 33 in SR2 and 10 in SR3. The pooled prevalence of clinical manifestations did not differ between sJIA and AOSD, except for myalgia, sore throat and weight loss, which were more frequent in AOSD than sJIA because they are likely ascertained incompletely in sJIA, especially in young children. Except for AA amyloidosis, more frequent in sJIA than AOSD, the prevalence of complications did not differ, nor did the prevalence of biological findings. Ferritin, S100 proteins and interleukin-18 (IL-18) were the most frequently used diagnostic biomarkers, with similar diagnostic performance. For MAS diagnosis, novel biomarkers such as IL-18, C-X-C motif ligand 9, adenosine deaminase 2 activity and activated T cells seemed promising.

Conclusion

Our results argue for a continuum between sJIA and AOSD.

PROSPERO registration number

CRD42022374240 and CRD42024534021.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Still’s disease was first described in children in 1897 and in adults 80 years later. However, in 1994, the paediatric form of the disease was given a different name, systemic juvenile idiopathic arthritis (sJIA), and since then most scientific studies have addressed sJIA and adult-onset Still’s disease (AOSD) separately.

• sJIA and AOSD classification criteria differ and therapeutic strategies have been defined independently for sJIA and AOSD, with important consequences in terms of clinical trial power, therapeutic strategy validation and knowledge dissemination.

WHAT THIS STUDY ADDS

• These systematic reviews and meta-analyses highlight the clinical and biological similarity of sJIA and AOSD in terms of prevalence of clinical and biological manifestations, including complications, as well as complications and biomarker diagnostic value.

• Three biomarkers seem relevant for both conditions: ferritin, S100 proteins (S100A8/A9 and A12) and interleukin-18 (IL-18). Additional evidence is needed to validate their robustness and make their use relevant in daily clinical practice.

• In addition to the standard laboratory measurements, novel biomarkers such as IL-18, interferon-γ-related biomarkers (C-X-C motif ligand 9 and adenosine deaminase 2 activity) and activated T cells are promising diagnostic biomarkers of macrophage activation syndrome.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The study results provide strong and robust arguments in favour of a continuum between sJIA and AOSD.

• They provide a strong rationale to produce joint recommendations for the diagnosis and management of sJIA and AOSD, as planned by the European Alliance of Associations for Rheumatology and the Paediatric Rheumatology European Society.

Introduction

Still’s disease was described in children in 1897 by Sir George Frederick Still and is today called systemic juvenile idiopathic arthritis (sJIA).¹ One century later, in the early 1970s, Eric Bywaters described a similar inflammatory condition in young adults, called adult-onset Still’s disease (AOSD).² Indeed, according to his study of 14 women aged 17–35 years, clinical features were identical to those of patients with sJIA. The cut-off of 16 years to discriminate between the two conditions was subsequently instituted but was arbitrary and related to the organisation of care structures (different inpatient units for children and adults with ‘rheumatism’ because of different care requirements).³

Many arguments support that sJIA and AOSD may correspond to the same rare non-familial (sporadic) systemic inflammatory disorder occurring at different ages.⁴ sJIA and AOSD are characterised by inappropriate activation of innate immunity, with typical features of autoinflammatory disease, particularly at onset, and by subsequent abnormalities in adaptive immunity during the chronic phase. Clinically, both entities are characterised by four cardinal symptoms: high fever, typically spiking and lasting several days or weeks; arthralgia and arthritis; skin rash, typically coloured salmon pink and macular or maculopapular; and increased leucocyte and neutrophil counts.^{5 6} Besides these cardinal features, the two conditions share many other clinical manifestations, including hepatomegaly, splenomegaly, lymphadenopathy and serositis. Furthermore, they exhibit common laboratory abnormalities, including increased erythrocyte sedimentation rate (ESR) and C reactive protein (CRP) level and hyperferritinaemia. In addition, the disease course and prognosis are comparable.^{5 6} Historically, the clinical course has been divided into three different phenotypes, described on the basis of the evolution of symptoms over time: monocyclic, polycyclic and chronic.⁴

For both sJIA and AOSD, a phenotypic dichotomy has been recognised: a more systemic inflammatory phenotype and a more chronic articular phenotype, with or without residual systemic features.⁴ Additionally, both are associated with a predisposition to life-threatening complications, such as macrophage activation syndrome (MAS), hepatitis and interstitial lung disease.^{5 6} MAS is a hyperinflammatory condition that should be suspected in patients with sJIA/AOSD with fever, cytopenia and hyperferritinaemia. Diagnostic scores (MAS/sJIA (MS) score or HScore)^{7 8} and classification criteria (2016 MAS criteria)⁹ are available to help clinicians in the early diagnosis. Early immunosuppressive treatment is associated with reduced mortality, both in adults and children. At present, MAS mortality remains approximately 10%,¹⁰ so identifying biomarkers related to the pathway involved in the development of MAS could be useful for targeted therapy and to improve management.

In addition, several data suggest that sJIA and AOSD are very similar in terms of the pathophysiology because the innate immune system plays a prominent role in both conditions.^{5 11 12} The overexpression of inflammatory cytokines, such as interleukin-1 (IL-1), IL-6, IL-18 and calcium-binding proteins, as well as the striking response to IL-1 and IL-6 inhibition have led to considering the conditions as complex, polygenic autoinflammatory syndromes.^{5 11 12} The finding of similar associations with human leucocyte antigen (HLA) alleles and cytokine gene polymorphisms indicates that sJIA and AOSD may be indistinguishable on a genetic level.¹³ While these patients do not exhibit classical features of autoimmunity, the adaptive immunity seems involved, particularly in patients with a chronic articular course.

According to the compelling evidence of their similarity, many experts believe that sJIA and AOSD are the same disease occurring at different ages.⁴ However, there has never been a formalised consensus, and most scientific studies have addressed sJIA and AOSD separately. As a result, the classification criteria used for clinical research purposes differ, and therapeutic strategies and clinical practice guidelines have been defined independently for sJIA and AOSD, thus reducing the power of clinical trials, limiting the dissemination of knowledge and hampering the implementation of optimal strategies in daily practice. Several Still’s disease experts have expressed the view that sJIA and AOSD should be considered a single disease, but evidence has been lacking for the case and to reach consensus.

The European Alliance of Associations for Rheumatology (EULAR) and the Paediatric Rheumatology European Society (PReS) decided to homogenise the perspective of rheumatologists of paediatric and adult disease and set up a task force to draft joint recommendations for the diagnosis and management of sJIA and AOSD. A systematic review (SR) of the literature was undertaken to inform the task force. The objectives of this study were to analyse the similarity of the diseases in terms of the prevalence of clinical manifestations (including complications) and the diagnostic value of biomarkers in both diseases. Finding that the diseases are similar in this SR would help demonstrate that sJIA and AOSD are the same disease occurring at different ages.

Methods

We performed three SRs, one on the prevalence of clinical and biological manifestations and complications (Prevalence SR1), another on the performance of biomarkers as diagnostic tests for sJIA and AOSD (Diagnostic SR2) and a third on the performance of biomarkers as diagnostic tests for MAS in both sJIA and AOSD (Diagnostic for MAS SR3). All three are reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.¹⁴

SR1: prevalence of manifestations

A review protocol was established a priori and registered in PROSPERO (CRD42022374240, https://www.crd.york.ac.uk/prospero).

Eligible studies

This first SR sought to include all cohort studies (retrospective or prospective) comparing sJIA to AOSD that described clinical and biological manifestations (including complications) and had a minimum of 20 patients per group. To be included, patients with sJIA had to meet the International League of Associations for Rheumatology (ILAR) criteria¹⁵ or at least one of the earlier historical classification criteria (American College of Rheumatology (ACR) 1972 criteria¹⁶ or Durban’s first revision of ILAR criteria¹⁷), or the Paediatric Rheumatology INternational Trials Organisation (PRINTO) criteria.¹⁸ Patients with AOSD had to meet Yamaguchi,¹⁹ Fautrel,²⁰ or Medsger and Christy²¹ classification criteria. We also retained studies that used criteria for at least one of the above-mentioned classifications, but the final diagnosis was from the physician.

Search strategy and selection process

Studies were identified by using sensitive search strategies in Medline (PubMed), Embase and Cochrane Library up to October 2022 (see online supplemental tables 1–3 for search strategies). Disease-related terms were used as search keywords, which involved a controlled vocabulary, specific MeSH headings and additional keywords. The references of included studies were also searched.

All records captured by the search strategies were downloaded to a bibliographic manager (EndNote), checked for duplicates, then screened for inclusion criteria by title and abstract by using Rayyan. Two independent reviewers (SM, SB) screened the records separately and compared selections. If a discrepancy occurred, consensus was reached by consulting a third reviewer (LC). In case of doubt, the article was selected for further review.

Data collection, outcomes and quality assessment

The articles selected were read in detail, and eligibility criteria were checked. One reviewer (SM) collected data in predesigned forms approved by all coauthors and crosschecked with a second reviewer (LC) in case of doubt. Variables to include were related to the study and design (type of study (cross-sectional, longitudinal, retrospective/prospective, other), country, sampling), the sample studied (disease (sJIA or AOSD), duration of evolution, evolution profile, ie, systemic/articular pattern, monocyclic or polycyclic and disease activity assessed by Pouchot score, original²² or modified by Rau et al²³), number of patients, age, sex, matching scheme, therapeutic strategies and the outcomes of interest.

The main outcomes of this review were clinical manifestations and their prevalence, biological findings and their prevalence, and complications and their prevalence (see online supplemental table 4 for details of each outcome). Additional outcomes were the different definitions used to define some manifestations such as fever, musculoskeletal or skin involvement, hepatomegaly, splenomegaly or lymphadenopathy (online supplemental table 4). For each descriptor, the absolute frequencies (n and N), or the mean and SD, depending on the data type, were extracted. In case of missing data or unusable data in an article, the corresponding author was contacted to obtain additional usable data.

The risk of bias of the included studies was assessed with an adapted form of the Hoy scale for prevalence studies²⁴ (online supplemental table 5).

Data analysis

We performed meta-analyses of proportions for the different qualitative descriptors by using the metaprop command in Stata. Proportions were pooled and displayed in tables and forest plots. CIs were based on the Wilson score procedure.²⁵ We tested whether the summary effect measure (prevalence) was equal to zero as well as heterogeneity (ie, whether the true effect in all studies was the same). Heterogeneity was quantified by using the I² measure.²⁶

SR2: diagnostic biomarkers for sJIA and AOSD

Here we sought to include studies reporting diagnostic biomarkers for sJIA and/or AOSD. We included only studies with data on accuracy (ie, reporting receiver operating characteristic (ROC) curve values, area under the ROC curve (AUC) and/or sensitivity and specificity). Medline (PubMed), Embase and Cochrane Library were searched up to 23 February 2023 (see online supplemental tables 7–9 for search strategies). Two independent reviewers (SM, ADM) screened records separately and compared selections. If a discrepancy occurred, consensus was reached by consulting a third reviewer (LC). In case of doubt, the article was selected for further review. The risk of bias of the included studies was assessed with the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2), which is designed to assess the quality of primary diagnostic accuracy studies.²⁷

SR3: diagnostic biomarkers for MAS

A review protocol was established a priori and registered in PROSPERO (CRD42024534021, https://www.crd.york.ac.uk/prospero). A study was eligible if patients had an MAS diagnosis by 2016 MAS classification criteria,⁹ the number of patients was ≥5 and it included diagnostic and/or prognostic information in terms of sensitivity, specificity and AUC. A biomarker was defined as prognostic if it was able to predict disease outcome (eg, death, need for intensive care, achievement of remission). Two independent reviewers (ADM and SM) screened records separately and compared selections. Medline (PubMed), Embase and Cochrane Library were searched up to 23 February 2023 by following the above-described methodology. The risk of bias was evaluated by using QUADAS-2 for diagnostic biomarkers and the Newcastle-Ottawa Scale for prognostic biomarkers. Case series were scored as high risk.

Results

SR1: prevalence of manifestations

The search identified 515 records. After removing 191 duplicates, 324 reports were screened, and one article, published immediately after the date of search, was included after a secondary search. Nine articles were assessed for detailed eligibility, and all but one were finally included (n=1010 participants). Online supplemental figure 1 describes the selection process.

Table 1 summarises the main characteristics of the included studies. Except for one cross-sectional study, the design of the studies was longitudinal; only one was prospective. The risk of bias was moderate in six studies and high in two (mean score 5 on the adapted Hoy scale for prevalence studies). In two studies, the authors clearly stated they were from referral centres; such information was missing for the others. The systemic/articular pattern was reported in only two studies, and the disease course was described in five. ILAR criteria were the most used for the diagnosis of sJIA (six studies, five using the 2001 second Edmonton revision and one the previous first 1997 Durban revision), followed by the ACR 1972 criteria (two studies) and PRINTO criteria (one study). Only three studies strictly applied the ILAR criteria for patient inclusion. Two studies used them as a diagnostic aid, but the final diagnosis was by physician opinion.^{28 29} For AOSD, the use of Yamaguchi criteria was more consensual: they were used in all eight studies. In addition to the Yamaguchi criteria, the Medsger and Christy and Fautrel criteria were used in one study each. As for sJIA, the final diagnosis of AOSD was retained by the physician in the same two studies. The Pouchot systemic score (original or modified by Rau et al) was available for four studies; disease activity was high.

Table 1. Main characteristics of the included studies in SR1: prevalence of manifestations.

First author, year, country	Study design, number of centres	Referral centre(s)	Number of sJIA/AOSD/total patients	Number of women with sJIA/AOSD/total patients (%)	Diagnostic criteria met, sJIA, %	Diagnostic criteria met, AOSD, %	Disease course	Pouchot systemic score	RoB: score, overall risk
Uppal, 1995,India30	LOB, retrospective,NR	NR	sJIA: 23AOSD: 31Total: 54	sJIA: 10 (43.4)AOSD: 14 (45.1)Total: 24 (44.4)	ACR 1972: 100	Yamaguchi: 100Medsger and Christy: 100	Monocyclic: 17.4%sJIA—22.6% AOSDPolycyclic: 17.4% sJIA—19.4% AOSDChronic: 52.2% sJIA—51.6% AOSDUnclassified: 13.0% sJIA—6.4% AOSD	Mean, original: 3.7 for both sJIA and AOSD	4
Lin, 2000, Taiwan32	LOB, retrospective, 1	NR	sJIA: 24AOSD: 21Total: 45	sJIA: 14 (58.3)AOSD: 12 (57.1)Total: 26 (57.7)	ACR 1972: 100	Yamaguchi: 100	Monocyclic: 66.7% sJIA—57.1% AOSDPolycyclic: 33.3% sJIA—42.9% AOSD	NR	5
Pay, 2006,Turkey31	LOB, retrospective, 6	Yes	sJIA: 25AOSD: 95Total: 120	sJIA: 13 (52.0)AOSD: 50 (52.6)Total: 63 (52.5)	ILAR criteria (Durban first revision): 100	Yamaguchi: 100	Monocyclic: 21.1% sJIA—21% AOSDPolycyclic: 41.7% sJIA—16.8% AOSDChronic: 29.2% sJIA—41.1% AOSDUnclassified: 8.3% sJIA—21.1% AOSD	NR	4
Inoue, 2016,Japan12	LOB, prospective, 1	NR	sJIA: 77AOSD: 33Total: 110	sJIA: 32 (41.5)AOSD: 21 (63.6)Total: 53 (48.2)	ILAR: 100	Yamaguchi: 100	NR	NR	5
Kudela, 2019,Germany28	Cross-sectional, retrospective, NR	NR	sJIA: 20AOSD: 30Total: 50	sJIA: 9 (45.0)AOSD: (20.0)Total: 15 (30.0)	ILAR: 55.5By physicians: 100	ILAR: 16.7Yamaguchi: 66.7By physicians: 100	NR	Mean, modified:3.5±1.1 for sJIA (n=15)4.0±1.4 for AOSD (n=19)	6
Ruscitti, 2017, Italy65	LOB, retrospective, 1	NR	sJIA: 21AOSD: 29Total: 50	sJIA: 15 (71.4)AOSD: 6 (20.7)Total: 21 (42.0)	ILAR: 100	Yamaguchi: 100	NR	Mean, original:4.620±2.110 for sJIA5.591±2.041 for AOSD	6
Ruscitti, 2022,Italy66	LOB, retrospective, 11	Yes	sJIA: 166AOSD: 194Total: 360	sJIA: 87 (52.4)AOSD: 92 (47.4)Total:179 (49.7)	ILAR: 100	Yamaguchi: 100	Monocyclic: 24.7% sJIA—34.5% AOSDPolycyclic: 45.2% sJIA—44.8% AOSDChronic: 21.7% sJIA—12.9% AOSD	Median, original:3 (2–5) for sJIA5 (4–7) for AOSD	5
Neau, 2022,France29	LOB, retrospective, 10	NR	sJIA: 86AOSD: 152Total: 238	sJIA: 48 (55.8)AOSD: 88 (57.9)Total: 136 (57.1)	ILAR: 50.7PRINTO: 55.5By physicians: 100	Yamaguchi: 60.8Fautrel: 65.7By physicians: 100	Chronic: 25.5% sJIA—21.7% AOSDUnclassified: 9.3% sJIA—10.5% AOSD	NR	5

The articles are sorted by year of publication.

The table reports for each study how the authors classified the course of the disease, although the strict definition of monocyclic/polycyclic/chronic and unclassified course may vary among studies (for the exact definition, please refer to each study). Disease pattern (systemic vs chronic articular) is not provided in the table because only 2 studies provided information on this issue: Lin 2000et al32 :(with 100% systemic pattern for both sJIA and AOSD), and Neau et al292022 (65.1% and 67.7% systemic pattern for sJIA and AOSD, respectively).

Pouchot systemic score, original22 or modified by Rau et al.23 Data are mean±SD or median [(interquartile rangeIQR]).

Risk of bias, adapted from Hoy scale. Summary of the overall risk of study bias: low risk (0–3), moderate risk (,4–5), high risk (.6–8). Green colour, low risk (no study had low risk in this table). Orange colour, moderate risk. Red colour, high risk.

All studies using the ILAR criteria were based on the second 2001 Edmonton revision,15 except, Pay et al 312006 which used the first 1997 Durban revision.17

For Inoue et al122016: the number of patients is 77 for sJIA, 33 for AOSD, 110 for total patients. However, for many clinical or biological features, the authors provide for each feature the number (n) reported on the number (N) of active patients, which is 66 for sJIA, 27 for AOSD, 93 in total.

For Kudela et al,28 the diagnosis was established by at least two experienced physicians.

For Neau et al, 29 the diagnosis was retained if it was confirmed by the referring physician at the last follow-up.

Note: tThis table does not include the age of patients because data were too heterogeneous: 3 studies gave the current age at time of study, 2 the mean age at onset and 4 the median age at onset or diagnosis. Similarly, it does not include the mean delay in diagnosis (because it was only provided in 2 studies), nor the mean follow-up (because it was only reported in 2 studies).

ACR 1972American College of Rheumatology 1972 criteriaAOSDadult-onset Still’s diseaseILARInternational League of Associations for RheumatologyLOBlongitudinal observationalNRnot reported or not clearly reportedPRINTOPaediatric Rheumatology INternational Trials OrganisationsJIAsystemic juvenile idiopathic arthritis

Table 2 summarises the main demographic characteristics of the samples as well as the patterns and clinical courses and criteria fulfilment by age group. Descriptors are provided as weighted average percentages or means. The female distribution was similar, about 50% in each group. The duration of follow-up was 6 months longer in the sJIA than AOSD group. The groups were comparable in distribution of patterns and courses.

Table 2. Demographic characteristics of the samples, patterns and courses, and criteria fulfilled, by group (SR1: prevalence of manifestations).

	Articles (n)	Total number of patients	Patients with sJIA (n)	Patients with AOSD (n)	sJIA	AOSD
Weighted average descriptor
% women	8	1010	431	579	53%	50%
Age (years)	3	193	107	86	8.5	46.5
Age at onset (years)*	2	99	47	52	5.1	27.2
Diagnostic delay (years)	2	99	47	52	5.2	4.4
Follow-up (months)	2	99	47	52	25.7	19.7
Weighted % descriptor
Systemic pattern†‡	2	283	110	173	73%	72%
Articular pattern§‡	1	238	86	152	26%	22%
Monocyclic course¶	4	579	238	341	28%	31%
Polycyclic course¶	4	579	238	341	41%	35%
Chronic course¶	4	772	300	472	26%	24%
Unclassified course¶	3	358	111	247	10%	14%
Weighted % criteria
ILAR (2001 Edmonton revision)**	5	372	342	30	87%	17%
ILAR (1997 Durban revision)**	1	25	25	–	100%	–
PRINTO	1	72	72	–	56%	–
Yamaguchi	8	570	–	570	–	88%
Fautrel	1	143	–	143	–	66%
ACR 1972	2	47	47	–	100%	–
Medsger and Christy	1	31	–	31	–	100%

Results are presented in weighted percentages or means.

^*Ruscitti et al 65 do not specify whether age is current age or age at onset or at diagnosis; however, the other parameters were collected at diagnosis, so we considered age to be age at diagnosis. Nevertheless, it could be current age, because 11 and 52 years seem too old for sJIA and AOSD onset, respectively.

^†In, Lin et al, 322000 100% of patients had initial systemic onset, but 46% of patients with sJIA and 38% of patients with AOSD exhibited chronic arthritis during follow-up.

^‡Only 2 studies provided information on the disease pattern (systemic vs articular): Lin et al322000: (with 100% systemic pattern at onset for both patients with sJIA and AOSD ) and Neau et al 29 (65.1% and 67.7% systemic pattern for sJIA and AOSD, respectively).

^§In, Neau et al29, 2022 the ‘articular pattern’ is defined as ‘chronic pattern’ and corresponded to the ‘chronic course’.

^¶The table reports for each study how the authors classified the course of the disease, although the strict definition of monocyclic/polycyclic/chronic and unclassified course may vary among studies (for the exact definition, please refer to each study).

^**All studies using the ILAR criteria were based on the second Edmonton revision,15 except, Pay et al 312006 which used the first Durban revision.17

ACR 1972American College of Rheumatology 1972 criteriaAOSDadult-onset Still’s diseaseILARInternational League of Associations for RheumatologyPRINTOPaediatric Rheumatology INternational Trials OrganisationsJIAsystemic juvenile idiopathic arthritis

The two groups did not differ in pooled prevalence of clinical manifestations, except for the three features, myalgia, sore throat and weight loss, which were more frequent in AOSD than sJIA (figure 1, online supplemental figure 2). Similarly, the groups did not differ in pooled prevalence of biological features, including leucocytosis, serum CRP level and ferritin level. The only exception was anaemia when defined by the cut-off of haemoglobin <120 g/L, which is more appropriate for adults. Indeed, during childhood, normal values change with age, and in several age groups, a value of 120 is well within normal limits (figure 2, online supplemental figure 3). Indeed, the difference was not evident when a threshold relevant for both children and adults was used (ie, haemoglobin <100 g/L). Other biological features were reported differently across studies, so obtaining a pooled mean was not possible.

Figure 1. Prevalence (pooled estimates) of clinical manifestations in patients with sJIA and AOSD (SR1: prevalence of manifestations). Values are pooled estimates of prevalence (95% CI). For each parameter, prevalence is represented in black for sJIA and grey for AOSD. ¹Heterogenity is summarised visually by asterisks: *low heterogeneity (I²<50%); **moderate heterogeneity (50<I²<75%); ***high heterogeneity (I²>75%). For details of the I² and p values of each group, please refer to the online supplemental material. ²Heterogeneity between groups is statistically different at p<0.05. For details of p values, please refer to the online supplemental material. The line "arthritis bis" gives the arthritis pooled estimate prevalence without the study by Inoue et al ¹² because it had a very low rate of arthritis in both age groups, accounting for the large part of variability when taken into account. sJIA, systemic juvenile idiopathic arthritis; AOSD, adult-onset Still’s disease.

Figure 2. Prevalence (pooled estimates) of modified biological features in sJIA and AOSD (SR1: prevalence of manifestations). Values are pooled estimates of prevalence (95% CI). For each parameter, prevalence is represented in black for sJIA and grey for AOSD. ¹Heterogenity is summarised visually by asterisks: *low heterogeneity (I²<50%); **moderate heterogeneity (50<I²<75%; there were no studies with moderate heterogeneity in this figure); ***high heterogeneity (I²>75%). For details of the I² and p values of each group, please refer to the online supplemental material. ²Heterogeneity between groups is statistically different at p<0.05. For details of p values, please refer to the online supplemental material. ANA, antinuclear antibody; AOSD, adult-onset Still’s disease; CRP, C reactive protein; ESR, erythrocyte sedimentation rate; Hb, haemoglobin; RF, rheumatoid factor; sJIA, systemic juvenile idiopathic arthritis.

Except for AA amyloidosis, the two groups did not differ in pooled prevalence of complications (figure 3, online supplemental figure 4). Although mortality was slightly higher in the AOSD than sJIA group (with high heterogeneity in the studies), the difference did not reach statistical difference. Regarding MAS, we found heterogeneity within groups, but not between groups. Of note, only one study reported on the frequency of thrombotic microangiopathy, tamponade, myocarditis or pulmonary hypertension, and, therefore, meta-analysis was not performed for these complications.²⁹ Two studies reported the frequency of amyloidosis. Although numbers were very low, AA amyloidosis was less frequent in AOSD than sJIA, and the difference between groups was statistically significant (p=0.009).

Figure 3. Prevalence (pooled estimates) of complications in patients with sJIA and AOSD (SR1: prevalence of manifestations). Values are pooled estimates of prevalence (95% CI). For each parameter, prevalence is represented in black for sJIA and grey for AOSD. ¹In Neau et al²⁹, the details of macrophage activation syndrome (MAS), thrombotic microangiopathy, tamponade, myocarditis and interstitial lung disease for each group (sJIA and AOSD) were obtained directly from the authors (not published data). ²Heterogenity is summarised visually by asterisks: *low heterogeneity (I²<50%); **moderate heterogeneity (50<I²<75%); ***high heterogeneity (I²>75%). For details of the I² and p values of each group, please refer to the online supplemental material. ³Heterogeneity between groups is statistically different at p<0.05. For details of p values, please refer to the online supplemental material. ⁴Note that for Neau et al [29], the authors reported a total n=26 in the published data, but in their Excel file n=24 (12 in sJIA group, 12 in AOSD group), so we considered n=24 for our meta-analysis. ⁵Only one study ²⁹ reported on the frequency of thrombotic microangiopathy, tamponade, myocarditis or pulmonary hypertension, and so meta-analysis was not performed for these complications. AOSD, adult-onset Still’s disease; sJIA, systemic juvenile idiopathic arthritis.

Table 3 provides the weighted frequency of articular involvement in sJIA and AOSD by joint. Only two papers provided the n/N.^{30 31} A third provided a sentence with the order of frequency of the involvement but without the n/N.³² The difference in frequency between groups was not present, except in the following joints: knees, ankle, elbow, hip and cervical spine, clinically higher in sJIA than AOSD.

Table 3. Weighted frequency of articular involvement in sJIA and AOSD by joint.

Joints involved	Articles (n)	Total number of patients	Patients with sJIA (n)	Patients with AOSD (n)	Weighted percentage topography sJIA	Weighted percentage topography AOSD
Knee	2	174	48	126	77	61
Wrist	2	174	48	126	69	69
Ankle	2	174	48	126	75	46
Elbow	2	174	48	126	50	36
MCP	2	174	48	126	27	29
PIP	2	174	48	126	33	29
DIP (hand)	2	174	48	126	4	3
Shoulder	2	174	48	126	27	25
MTP	2	174	48	126	19	11
Hip	2	174	48	126	23	6
Cervical spine	1	120	25	95	24	1
Temporomandibular joint	1	120	25	95	4	3

In bold, joints in which the difference between groups was relevant.

AOSDadult-onset Still’s diseaseDIPdistal interphalangealMCPmetacarpophalangealMTPmetatarsophalangealPIPproximal interphalangealsJIAsystemic juvenile idiopathic arthritis

The weighted percentages of treatments used in included studies are found in online supplemental table 6. Non-steroidal anti-inflammatory drugs were more used in sJIA, and glucocorticoids and conventional synthetic disease-modifying antirheumatic drugs (DMARDs; methotrexate) were more used in AOSD. A greater percentage of children received biologic DMARDs than adults.

SR2: diagnostic biomarkers for sJIA and AOSD

The search identified 1099 records. After removing 120 duplicates, 972 reports were screened, 118 were assessed for eligibility and 33 were finally included. Online supplemental figure 5 describes the selection process.

The SR retrieved many different potential diagnostic biomarkers (online supplemental figure 6 and online supplemental table 10). Only three were analysed in several studies of both sJIA and AOSD: ferritin/glycosylated ferritin, the S100 protein group (S100A8/A9/A12) and IL-18. Table 4 summarises each of these studies and the cut-offs for biomarkers and accuracy data. Online supplemental table 11 provides the QUADAS-2 evaluation of risk of bias of the retrieved studies; most were at moderate to high risk of bias. We did not pool the results in a meta-analysis because of the variability in populations compared and the definition of a gold standard.

Table 4. Main serum diagnostic biomarkers in sJIA and AOSD and their performance (SR2: diagnostic biomarkers for sJIA and AOSD).

Biomarker	Cases (n)	Controls (n)	Cut-off(ng/mL)*	Sensitivity (%)	Specificity (%)	AUC	Ref
Ferritin
Ferritin	sJIA (11)	Fever of unknown origin† (29)	>5N	91.0	48.0	NA	67
Ferritin	sJIA (102)	Fever related to infections (60), autoinflammatory diseases‡ (97) or others§ (98)	178.5	84.4	69.1	0.840	68
Ferritin	AOSD (49)	Infections¶ (21), liver disease** (22), other systemic diseases†† (62), fever of unknown origin (5) and neoplasia‡‡ (10)	>N>5N	67.340.8	35.880.0	NANA	69
Ferritin	AOSD (73)AOSD (37)	Test cohort: sepsis (56)Validation cohort: sepsis (28)	1120.01120.0	74.596.7	94.150.0	0.8820.783	70
Ferritin	AOSD (36)	Rheumatoid arthritis (40) and healthy controls (33)	273.3	66.7	97.0	0.828	71
Ferritin	AOSD (42)	Sepsis and ANCA-vasculitis (46)	819.0	79.6	71.4	NA	72
Ferritin	AOSD (82)	Sepsis (48)	1120.0	74.7	88.9	0.887	73
Ferritin	AOSD (68)	Sepsis (55)	1086.4	73.8	90.0	0.872	74
Ferritin	AOSD (36)	Patients with ferritin level >1000 ng/mL but not fulfilling the Yamaguchi criteria (144)§§§§	2500.05000.010 000.0	74.062.045.0	73.092.097.0	0.7300.7700.710	75
Ferritin	AOSD (53)	Healthy controls (60)	759.9	100	84.3	0.940	76
Glycosylated ferritin (GF)
GF	AOSD (49)	Infections¶ 21), liver disease** (22), other systemic diseases†† (62), fever of unknown origin (5) and neoplasia‡‡ (10)	≤20%	79.5	66.4	NA	69
GF	AOSD (28)	Immune-mediated inflammatory diseases (70), infectious diseases (51), haematological malignancies (24), solid cancers (11), acute hepatitis (11), other¶¶ (36)	≤16%	88.5	63.2	0.794	77
Ferritin and GF
Ferritin and GF	sJIA (11)	Fever of unknown origin† (29)	F>5N+GF≤20%	81.8	48.3	NA	67
Ferritin and GF	AOSD (49)	Infections¶ (21), liver disease** (22), other systemic diseases†† (62), fever of unknown origin (5) and neoplasia‡‡ (10)	F>N+GF≤20%F>5N+GF≤20%	70.543.2	83.292.9	NANA	69
S100A8 protein (also called MRP8)
S100A8 protein	sJIA (23)	Acute lymphoblastic leukaemia (20)Kawasaki disease (26)Severe infection (18)	115.0362.0NA	95.257.1NA	78.9100NA	0.8950.6970.549	78
S100A9 protein (also called MRP14)
S100A9 proteins	sJIA (23)	Acute lymphoblastic leukaemia (20)Kawasaki disease (26)Severe infection (18)	15.3NANA	100NANA	100NANA	0.9740.5400.654	78
S100A8/S100A9 protein heterocomplex (also called MRP8/14 heterocomplex or calprotectin)
S100A8/A9	sJIA (21)	AID*** (8) + SURFS††† (14)	5630.0	74.0	91.0	0.900	79
S100A8/A9	sJIA (60)	Systemic infections (85)	9200.0	95.0	95.0	0.747	80
S100A8/A9	sJIA (59)sJIA (102)	Cohort A: untreated fever among infections (34), autoinflammatory diseases‡ (55) and miscellaneous§ (47)Cohort A+B: untreated fever among infections (60), autoinflammatory diseases‡ (97) and miscellaneous§ (98)	9160.0‡‡‡30 650.014 650.0¶¶¶30 650.014 650.0¶¶¶	72.974.678.079.484.3	90.488.283.189.080.8	0.8710.8740.8690.9100.900	68
S100A8/A9	AOSD (46)	Rheumatoid arthritis (34), primary Sjögren’s syndrome (40), systemic lupus erythematosus (39), osteoarthritis (20), healthy controls (49)	45 488.0	63.0	80.1	NA	81
S100A8/A9	AOSD (36)	Rheumatoid arthritis (40) and healthy controls (33)	4550.0	69.4	98.0	0.845	71
S100A12 proteins
S100A12	sJIA (21)	AID*** (8) + SURFS††† (14)	363.0	71.0	89.0	0.900	79
S100A12	sJIA (60)	Systemic infections (83)NOMID (18)Muckle-Wells syndrome (17)Acute lymphoblastic leukaemia (40)Acute myeloblastic leukaemia (5)Healthy controls (45)	80080010001000150150	85.084.078.078.097.098.0	89.072.0100100100100	0.8810.8860.9720.9811.0000.994	82
Interleukin (IL)-18††††
IL-18	sJIA (20)	Rheumatic and/or inflammatory diseases****(23)	11 473.5	61.0	100	0.774	28
IL-18	sJIA (102)	Untreated fever among infections (60), autoinflammatory diseases‡ (97) and miscellaneous§ (98)	148.7	70.4	82.9	0.785	68
IL-18	sJIA (23)	Acute lymphoblastic leukaemia (20)Kawasaki disease (26)Severe infections (18)Non-systemic JIA (18)	40 600.020 800.020 000.016 800.0	95.2100100100	100100100100	0.9951.0001.0001.000	78
IL-18	AOSD (23)	Severe COVID-19 (55)	190.5	95.8	91.3	0.948	83
IL-18	AOSD (26) (active)	Healthy controls (21)Rheumatoid arthritis (21)Systemic sclerosis (21)Systemic lupus erythematosus (21)	312737766366	86.780.995.270.0	61.546.146.161.5	0.7010.5860.5650.640	84
IL-18	AOSD (36)	Rheumatoid arthritis (40) and healthy controls (33)	366.1	91.7	99.1	0.983	71
IL-18	AOSD (70)	Sepsis (22)	543.0	93.7	83.3	0.884	85
IL-18	AOSD (30)	Rheumatic, autoinflammatory, infectious and/or haematological diseases (65)	832.5	80.0	81.5	0.870	28
IL-18	AOSD (16)	Febrile diseases (25), among them HLH (5)	20 000.0	94.0	96.0	NA	86
IL-18	AOSD (39)	Sepsis (18)	148.9	78.3	88.6	0.864	87
IL-18	AOSD (46, of which 9 with MAS)	Adult HLH (31)	18 550.0	90.3	93.5	0.910	88
IL-18	AOSD (17)	Sepsis (37)	5252.0	94.1	96.1	0.940	89

All the biomarkers reported were tested in blood (serum or plasma). For each biomarker, the articles are listed in alphabetical order for each entity (sJIA and AOSD).

^*For each biomarker, units are specified in the column: ng/mL (=µg/L) for ferritin, % for glycosylated ferritin, ng/mL for S100 proteins and IL-18.

^†Fever of unknown origin included other rheumatic diseases (Kawasaki disease, systemic lyupus erythematosus, urticarial vasculitis, poliyarteritis nodosa, Ttumour Nnecrosis Ffactor Rreceptor A-associated Pperiodic Ssyndrome), systemic infections, haemophagocytic lymphohistiocytosis, neoplasia (myeloid leukaemia, inflammatory myofibroblastic tumour), miscellaneous (insipidus diabetes, antibiotic-induced fever, carbamazepine-induced fever) and unidentified origin.

^‡Autoinflammatory diseases (AIDs) included TRAPS (Tumour necrosis factor receptor-associated periodic syndrome), CAPS (Cryopyrin-associated periodic syndrome), HIDS (Hyperimmunoglobulin D syndrome, also known as Mevalonate kinase deficiency or MKD), TRAPS+CAPS, PFAPA syndrome and undifferentiated AIDs (familial Mediterranean fever (FMF) excluded).

^§Miscellaneous included Behçet’s disease, connective tissue diseases, CRMO (chronic recurrent multifocal osteomyelitis), haematological/oncological diseases, non-systemic JIA, reactive arthritis, vasculitis, other/unknown diagnoses.

^¶Infections included bacterial (pyogenic bacteria or mycobacteria), viral (human immunodeficiency virusHIV, hepatitis A virus, hepatitis B virus, EBV (Epstein Barr virus), parvovirus B19), or parasitic (toxoplasmosis); all patients with viral hepatitis had predominantly systemic manifestations. One patient with viral hepatitis had also haemophagocytic lymphohyistiocytosis.

^**Twenty-two patients had a specific liver disease, notably Gaucher’s disease or haemochromatosis.

^††Systemic diseases included unclassified polyarthritis, polyarteritis nodosa and other vasculitis, RA, SLE, sarcoidosis, giant cell arteritis or polymyalgia rheumatica, SpA, relapsing polychondritis, familial Mediterranean fever, polymyositis, scleroderma, or other inflammatory conditions.

^‡‡Neoplasia included non-Hodgkin’s lymphoma, acute leukaemia, lung cancer, Hodgkin’s disease, and prostate cancer.

^§§Control patients were matched 1:4 for age and sex with patients with AOSD (but no further precision on the diseases presented by the controls).

^¶¶The ‘other’ group included crystal arthropathies (n=5), urticaria (n=8) and various conditions (n=23, DRESS (Drug Reaction with Eosinophilia and Systemic) syndrome, osteoarthritis, Dressler syndrome, thrombotic microangiopathy, venous thromboembolic disease, fever of unknown origin, fibromyalgia, focal and segmental glomerulosclerosis, myocardial infarction, subacute granulomatous thyroiditis, alcohol use disorder).

^***The AID(autoinflammatory disease (AID) group) included patients with defined genetic or clinically diagnosed fever syndromes other than sJIA. This group consists of patients with (FMF),; Ttumour Nnecrosis Ffactor Rreceptor A-associated Pperiodic Ssyndrome,; Pperiodic Ffever, Aaphthous Sstomatitis, Ppharyngitis and Aadenitis syndrome; and Muckle-Wells syndrome.

^†††Patients with periodic or recurrent patterns of fever but without a defined disease were classified as having systemic undifferentiated recurring fever syndromes (SURFS).

^‡‡‡With experimental ELISA. With Bühlmann MRP8/14 ELISA. With Quantum Blue sCAL test.

^§§§With Bühlmann MRP8/14 ELISA.

^¶¶¶With Quantum Blue sCAL test.

^****There were 22 different rheumatic and/or inflammatory diseases.

^††††IL-18 is present in the blood as a free molecule or bound to IL-18 -binding protein (IL-18BP), total IL-18 being a combination of both. The studies identified did not clearly specify which form of IL-18 they were referring to, and therefore we inferred that they were referring to total IL-18.

AIDAutoinflammatory diseaseANCAantineutrophilic cytoplasmic antibodyAOSDadult-onset Still’s diseaseAUCarea under the receiver operating characteristic curveCAPSCryopyrin-associated periodic syndrome CRMOChronic Recurrent Multifocal OsteomyelitisDRESSDrug Reaction with Eosinophilia and Systemic SymptomsEBVEpstein Barr virusFerritin>5Nferritin level >5 times the normal levelFerritin>Nferritin level above the normal valuesHIDShyperimmunoglobulin-D syndromeHLHhaemophagocytic lymphohistiocytosisILinterleukinMASmacrophage activation syndromeMKDMevalonate kinase deficiencyNAdata not availableNOMIDneonatal-onset multisystem inflammatory diseasePFAPAperiodic fever, aphthous stomatitis, pharyngitis, adenitisRArheumatoid arthritissJIAsystemic juvenile idiopathic arthritisSLEsystemic lupus erythematosusTRAPSTumor necrosis factor receptor-associated periodic syndrome

SR3: diagnostic biomarkers for MAS

From the 979 articles retrieved, we finally selected 10 articles that fulfilled the eligibility criteria described (online supplemental figure 7). Eight articles concerned sJIA and two AOSD. Among the biomarkers identified, some were ‘classical’ (ie, standard laboratory measurements) and others ‘novel’ (ie, non-standard laboratory measurements of inflammation related to pathogenic pathways of MAS). High sensitivity/specificity was defined as values >70% and moderate sensitivity/specificity as 50–70%.

Classical biomarkers for MAS

Seven articles evaluated the classical biomarkers of MAS33,39: two studies included adults, and five studies included children. Online supplemental table 12 summarises the main data and the risk of bias for each study. Two studies before the development of the 2016 MAS classification criteria were included because from an analysis of the articles we could establish that the patients satisfied the 2016 MAS criteria.

Ferritin level was the most frequently reported diagnostic biomarker for MAS. It showed high sensitivity but moderate specificity to differentiate sJIA/AOSD-related MAS and active sJIA/AOSD without MAS or infections. The cut-off values differed depending on the population (adult or paediatric) and the controls used. The ratio or ferritin level to ESR showed high sensitivity and specificity to differentiate sJIA-related MAS and infections. Only one study evaluated the glycosylated ferritin fraction in AOSD-related MAS versus AOSD without MAS; it showed very high sensitivity but low specificity.

CRP level was tested as a diagnostic biomarker to differentiate sJIA-related MAS and virus-associated or familial haemophagocytic lymphohistiocytosis (HLH) but showed moderate sensitivity and specificity. It was the only biomarker tested as a prognostic biomarker for MAS in one study in which high CRP level predicted mortality in AOSD-related MAS versus AOSD without MAS, with moderate sensitivity and low specificity. Other inflammatory/tissue damage biomarkers (albumin, aspartate aminotransferase, soluble CD25, fibrinogen, lactate dehydrogenase, neutrophil, platelet count and white cell count) were evaluated as diagnostic biomarkers for sJIA-related MAS versus active sJIA or other forms of HLH. All showed moderate sensitivity and specificity.

Novel biomarkers for MAS

Four studies investigated the accuracy of novel biomarkers as potentially diagnostic for MAS.3340,42 Table 5 summarises their main data and risk of bias. These studies included only paediatric patients with sJIA-related MAS. Patients with sJIA without MAS or with other (primary or secondary) forms of HLH were used as control group.

Table 5. Novel diagnostic biomarkers for MAS: characteristics of the studies and parameters of accuracy (SR3: diagnostic biomarkers for MAS).

Serum biomarker	Case (n)	Controls (n)	Cut-off	Sensitivity (%)	Specificity (%)	AUC	First author (year)	RoB*
Total IL-18†	sJIA-MAS (23)	sJIA (65)	>68 363 ng/mL	95.0	90.0	0.974	Lee (2020)33
Total IL-18†	sJIA-MAS (11)	pHLH (10)	1157 pg/mL	85.7	67.5	0.822	Kessel (2021)40
		sHLH (12)	1022 pg/mL	73.2	77.5	0.787
IL-18/CXCL9	sJIA-MAS (11)	pHLH (10)	4.5	82.1	75.0	0.794	Kessel (2021)40
		sHLH (12)	4.4	68.3	77.5	0.719
IL-18/CXCL10	sJIA-MAS (11)	pHLH (10)	14.8	78.6	70.0	0.745	Kessel (2021)40
		sHLH (12)	10.6	65.9	72.5	0.714
S100A12	sJIA-MAS (11)	pHLH (10)	535 ng/mL	96.0	91.4	0.938	Kessel (2021)40
		sHLH (12)	635 ng/mL	83.8	91.4	0.847
S100A12/CXCL9	sJIA-MAS (11)	pHLH (10)	1593	88.0	91.4	0.907	Kessel (2021)40
		sHLH (12)	2231	81.1	85.7	0.827
S100A12/CXCL10	sJIA-MAS (11)	pHLH (10)	9293	92.0	85.7	0.918	Kessel (2021)40
		sHLH (12)	10 959	81.1	85.7	0.846
ADA2 activity	sJIA-MAS (23)	sJIA (65)	>25.7 U/L	91.0	95.0	0.939	Lee (2020)33
CD38high/HLA-DR+CD8+ T cells	sJIA-MAS (14)	Active sJIA (27)	>12.05%	100	85.2	0.960	De Matteis (2022)41
CD4dimCD8+ T cells	sJIA-MAS (14)	Active sJIA (27)	>0.85%	92.9	81.5	0.900
CXCL9	sJIA-MAS (23)	sJIA (65)	>203 ng/mL	90.0	77.0	0.882	Lee (2020)33
sTNFR-I/sTNFR-II	sJIA-MAS on TCZ (36)	sJIA on TCZ (16)	>4.71	100	83.0	0.972	Irabu (2020)42

^*Risk of bias: green=low (no study showed a low risk in this table), yellow=intermediate, red=high. **IL-18 is present in the blood as a free molecule or bound to IL-18 binding protein (IL-18BP), total IL-18 being a combination of both.

^†IL-18 is present in the blood as a free molecule or bound to IL-18-binding protein (IL-18BP), total IL-18 being a combination of both.

ADA2adenosine deaminase 2AUCarea under the receiver operating characteristic curveCXCL9C-X-C motif ligand 9CXCL10C-X-C motif ligand 10HLAhuman leucocyte antigenIL-18interleukin-18MASmacrophage activation syndromepHLHprimary haemophagocytic lymphohistiocytosisS100A12S100 calcium-binding protein A12sHLHsecondary haemophagocytic lymphohistiocytosissJIAsystemic juvenile idiopathic arthritissTNFRserum tumour necrosis factor receptorTCZtocilizumab

Total IL-18 level was used as a diagnostic biomarker of MAS alone or combined with C-X-C motif ligand (CXCL9 or CXCL10). High levels of total IL-18 had high sensitivity and specificity for distinguishing sJIA-related MAS from sJIA without MAS but not the other forms of secondary HLH (sHLH). Increased IL-18/CXCL9 and IL-18/CXCL10 ratios had moderate sensitivity and specificity for distinguishing sJIA-related MAS from sHLH but with good sensitivity and specificity for distinguishing sJIA-MAS from primary HLH (pHLH). S100A12, alone or combined with CXCL9 or CXCL10, had high sensitivity and specificity for distinguishing sJIA-related MAS from pHLH or sHLH. CXCL9 alone (ie, not combined with IL-18 or S100A12) yielded high sensitivity and specificity for distinguishing sJIA-MAS from sJIA without MAS. High levels of adenosine deaminase 2 (ADA2) activity presented high sensitivity and specificity for distinguishing sJIA-related MAS from sJIA without MAS. High CD38^high/HLA-DR+CD8+ T-cell count and CD4^dimCD8+ T-cell count distinguished sJIA-MAS from active sJIA without MAS with high sensitivity and specificity. Finally, a high TNFR-I/TNFR-II ratio showed high sensitivity and moderate specificity for differentiating patients with sJIA-MAS and patients with sJIA receiving tocilizumab.

Discussion

To our knowledge, this study is the first SR of comparative cohorts and meta-analysis examining the prevalence of clinical manifestations (including complications) and laboratory findings in sJIA and AOSD. We found a globally similar prevalence of clinical and biological manifestations for sJIA and AOSD, including complications. This is a strong argument supporting a continuum between the two entities.

The differences in prevalence were few and not substantial and can be easily explained by confounding factors. Myalgia and sore throat were more frequent in adults than children. These are symptoms that are difficult for young children to report, leading to a possible reporting bias. Furthermore, sore throat is one of the items required in the Yamaguchi criteria and thus recorded/noted by rheumatologists caring for adults but is not present in ILAR criteria; therefore, paediatricians may be less likely to look for it,^{4 43} while it is not in the criteria used in the paediatric age. Weight loss was more frequently mentioned in adults than children, who rather have a growth curve break. The conditions did not differ in the frequency of anaemia with a haemoglobin cut-off of <100 g/L, which was more appropriate, because children physiologically have lower levels of haemoglobin compared with adults.⁴⁴ The authors of these comparative studies should probably have used different standards for this factor for paediatric and adult populations. The same remark can be applied to leucocytosis; however, no statistically significant difference was noted whatever the threshold. Children have more prolonged exposure to the uncontrolled inflammatory process than do adults, because the disease starts earlier in life.⁴⁵ However, the two studies reporting AA amyloidosis were older (1995 and 2006),^{30 31} performed prior to the widespread use of biologics, which may have greatly reduced the prevalence of AA amyloidosis in recent years.^{46 47} Notably, AA amyloidosis was not reported in the retrieved articles that were more recent than 2006. One last difference is a potential different distribution of the affected joints between children and adults (table 3). The small number of studies (two papers) requires caution in generalising these results. Other comparative series have shown a higher similarity in the topography of affected joints between sJIA and AOSD.⁴⁸ This observation seems consistent with the results of a third study we retrieved,³² in which the authors reported that ‘the affected joints were the knees, ankles, wrists, elbows, proximal interphalangeal joints and metacarpophalangeal joints (in order of frequency) in children and adults alike’. However, this study could not be included in our meta-analysis because the n/N for each joint involvement were not provided.

We did not find any difference between sJIA and AOSD in terms of sex in that the weighted average of women was about 50% in each group. Although the sex ratio is 1:1 in the paediatric-onset forms, some authors have suggested that the adult form may more frequently affect women (70% vs 30%).⁴ However, these data were mostly from rheumatological series with small numbers (61–104 patients) and included mainly chronic joint forms. More recent series have reported a more balanced sex ratio, although the number of patients is still limited.^{49 50}

Although many paediatricians often consider ILAR 2001 criteria too stringent because they require at least one joint with arthritis persisting for at least 6 weeks, we found a globally similar prevalence of arthritis between sJIA and AOSD in our meta-analysis. This observation could be explained in part by only three studies strictly applying the ILAR 2001 criteria for including paediatric patients, and one applied the previous 1997 Durban revision of the ILAR criteria, which are looser because they take into account the possibility of developing arthritis only after 6 months of systemic illness. In fact, the need for arthritis, which is mandatory in the ILAR 2001 classification, is debated among paediatricians because many young patients do not have arthritis (but rather arthralgia) and yet have all the other classic manifestations of Still’s disease.⁴³ Thus, a revision of the classification criteria for the paediatric form has recently been proposed,¹⁸ and the presence of persistent arthritis may no longer be considered mandatory in the future.

Concluding that sJIA and AOSD are the same disease occurring at different ages of life would eventually imply the need to use the same classification criteria. Indeed, in the current configuration, rheumatologists of children and adults may use different or even contradictory terms when referring to the same disease.⁵¹ Indeed, a common disease name and classification criteria would allow for homogenising the diagnosis and management of paediatric and adult patients and would open the door to joint clinical research and trials. The question remaining is whether new common criteria should be established or whether already existing classification criteria can be used interchangeably in children and adults. For instance, the Yamaguchi criteria may be useful for classifying sJIA.⁵²

Our SR retrieved many different potential diagnostic biomarkers (online supplemental figure 6 and online supplemental table 10). However, only three were analysed in several studies in both sJIA and AOSD: ferritin/glycosylated ferritin, the protein S100 group (S100A8/A9/A12) and IL-18. The controls used between the different studies were very different, so generalising results is impossible. Therefore, sensitivity, specificity, cut-offs and AUC values provided in table 4 must be interpreted with caution. Nevertheless, these diagnostic biomarkers look promising, and research should start using them in a more standardised way. One potential limitation of our SR on diagnostic biomarkers is that we included only studies that had accuracy data (ROC, AUC, sensitivity and specificity values). Hence, it misses more descriptive studies with only correlations, but this would have led to a high number of studies and lower scientific pertinence. Other biomarkers have been proposed but only in small cohorts, without replication in another population (online supplemental figure 6 and online supplemental table 10).

The same limitations (accuracy parameters, controls and missing descriptive studies with only correlations) apply to SR3 on diagnostic biomarkers for MAS (table 5 and online supplemental table 12). MAS occurring during sJIA and AOSD is a hyperinflammatory condition with significant morbidity and mortality. Early recognition and diagnosis can be challenging, and remission is often achieved with the association of multiple therapies. Clinical scores and classification criteria have recently been developed to help clinicians diagnose MAS. In sJIA, the 2016 classification criteria for MAS have shown high specificity (99%) and sensitivity (73%).⁹ In patients with AOSD, these criteria were useful to identify MAS with high specificity (100%) and sensitivity (70%) and helped identify patients at high risk of mortality.^{53 54} Two diagnostic scores for MAS have been developed, the MS score⁷ and the HScore⁸: they have high sensitivity (both 91.3%) and specificity (83.8% and 90.2%).⁵⁵ The MS score was developed in sJIA and the HScore was developed in adults with various forms of sHLH, which included only a few patients with AOSD. All these scores and criteria include variable combinations of elevated ferritin levels, inappropriately low cell counts, elevated liver function test results and signs of intravascular activation of coagulation, all known to be part of the clinical and laboratory pattern associated with hyperinflammation.⁵⁶ Our SR3 yielded several papers confirming ferritin level as a diagnostic biomarker for MAS with high sensitivity. However, we found scarce data, or no data, concerning the other laboratory factors (CRP, albumin, aspartate aminotransferase, soluble CD25, fibrinogen, lactate dehydrogenase, neutrophil, platelet count and white cell count), which are part of these criteria and scores. To improve the early recognition of MAS, studies examining the predictive value of each of these laboratory measurements are warranted. In addition, validation of the MAS 2016 criteria in populations of different ages and the establishment of cut-offs for age are needed.

With the progressive recent understanding of the pathogenesis of MAS, novel diagnostic biomarkers have been investigated to improve the diagnosis of MAS (ie, with earlier identification with higher sensitivity and specificity). Our SR3 retrieved only paediatric studies of these potential novel biomarkers for MAS. No studies involving adults could be linked to our inclusion criteria, which required diagnosis by the MAS 2016 classification criteria, which are more frequently used by paediatricians than physicians for adults. Therefore, we may have missed some studies of AOSD-related MAS using other criteria. Also, although MAS 2016 criteria are the most rigorous to classify MAS, they have been recently developed and so would appear to bias against older studies. In any case, the first novel biomarker identified is IL-18. It is an interferon-γ (IFN-γ)-inducing cytokine that has been involved in the pathogenesis of MAS in the context of sJIA/AOSD. It is present in the blood as a free molecule or bound to IL-18-binding protein (IL-18BP), total IL-18 being a combination of both.⁵⁷ Free IL-18 is biologically active. Markedly elevated levels of total IL-18 are suggestive of predisposition to the development of MAS in the sJIA population.^{58 59} Elevated total IL-18 level seems to distinguish MAS from active sJIA with high sensitivity and specificity.³³ One study reported significantly higher free IL-18 levels in sJIA-MAS than sJIA without MAS and other forms of HLH, with no relevant differences in level of IL-18BP between MAS and other forms of HLH.⁵⁷ Because the assay for free IL-18 is not widely available and levels of free IL-18 are highly related total IL-18 level,⁵⁷ measurement of total IL-18 is routine, and clinically graded tests are becoming available. Consistent with the activation of the IFN-γ pathway, high CXCL9 level and ADA2 activity, both induced by IFN-γ, have been reported in MAS in both children and adults.^{33 57 60 61} Indeed, CXCL9 level and ADA2 activity also seem good diagnostic biomarkers able to differentiate MAS and active sJIA, with high sensitivity and specificity (table 5). Furthermore, ADA2 activity is highly related to total ADA activity, whose detection is more convenient.⁶² Hence, total ADA activity may be a MAS biomarker.⁶² In addition, serum levels of total IL-18 and S100A12 differentiate MAS and other forms of HLH (primary and secondary) with high sensitivity and specificity.⁴⁰ This may be particularly helpful in the context of MAS presenting at the onset of Still’s disease when the diagnosis of the disease itself is not yet clear. Markedly elevated populations of activated CD8 T cells (CD38^high/HLA-DR+CD8+ T cells and CD4^dimCD8+ T cells) are present in MAS versus active sJIA.⁴¹ CD8 T lymphocytes have been found pathogenic in some HLH models, and their count is elevated in primary (p)HLH and infection-associated HLH.^{63 64} Activated CD8 T lymphocytes, identified by CD38^high/HLA-DR+ positivity, are also able to distinguish HLH (primary and infection associated) from sepsis.⁶⁴ They appear to differentiate MAS and active sJIA with high specificity and sensitivity. Both populations are similarly increased in all forms of sHLH, which demonstrates that they do not differentiate MAS in Still’s disease and other forms of sHLH.

Overall, these observations with the various potential diagnostic biomarkers are reported in a limited number of studies, and therefore, confirmation in larger series is needed. Moreover, to allow widespread use of these novel biomarkers in clinical practice, cross-validation between different laboratories is needed as is the identification of standardised international cut-off values. An international project involving North American and European centres is ongoing. No data about prognostic biomarkers have been reported, except one study reporting high CRP level as a predictor of mortality in MAS during AOSD, with high sensitivity and low specificity.³⁴ Taking into account the previous considerations and their increasing availability, IFN-γ-related biomarkers (CXCL9 and IL-18) seem to be novel biomarkers that could improve the diagnosis and management of MAS. Moreover, activated T cells are also promising because they can be easily evaluated with a simple flow cytometry protocol and seem to help in the differential diagnosis of MAS versus active sJIA/AOSD but also many forms of ppHLH and sHLH.

Our study has some limitations. The studies retrieved were of medium to low quality (medium to high risk of bias), which is inherent to studies of rare diseases and retrospective study designs (SR1 on comparative cohorts) and diagnostic biomarker studies (SR2 and SR3). In SR1, we included only cohorts that compared children and adults. Including all single cohorts would have been cumbersome and could have been a source of bias in reporting (reporting is more homogeneous in comparative cohorts). We chose the cut-off of ≥20 patients in each group for two reasons: sJIA/AOSD is a rare disease, so cohorts comparing children versus adults with a high number of patients in each group are rare; it would have been difficult to have any prevalence below at least 20 patients in each group. Of note, we had originally planned to collect information on biomarkers in these comparative studies. Unfortunately, the data were almost non-existent in these studies and therefore not usable, which justified the second SR specifically dedicated to diagnostic biomarkers and focusing on single cohorts. We also initially planned to examine the different definitions used to define some manifestations such as fever and musculoskeletal and skin involvement because these items often lack a consensual definition in the literature. Unfortunately, most of the selected articles did not provide definitions of these features, so we could not meet our objective. Complications were reported only by a few studies. Some (thrombotic microangiopathy, tamponade, myocarditis or pulmonary hypertension) were reported in only one study, so meta-analysis could not be performed.

Our data provide strong and robust arguments favouring the similarity of sJIA and AOSD (ie, a continuum between the two entities). They validate the rationale to produce joint recommendations for the diagnosis and management of sJIA and AOSD, as planned by EULAR and PReS. It is the first step for future work aimed at proposing a unique name for sJIA and AOSD, as well as a common and consensual set of classification criteria. The present work will also open the way to common research to validate diagnostic and prognostic biomarkers and establish new therapeutic strategies for the disease and its complications.

Protocol and registration

The protocol for SR1 was registered in PROSPERO (CRD42022374240, https://www.crd.york.ac.uk/prospero).

In the initial protocol, we had planned to include only studies whose participants met the ILAR criteria for patients with sJIA and the Yamaguchi or Fautrel criteria for patients with AOSD. However, because some of the selected studies were published before 2004, when the ILAR criteria were revised for the second time, the protocol was secondarily amended to include studies that used at least one of the earlier historical classification criteria (ACR 1972 or Durban’s first revision of ILAR criteria for sJIA; Medsger and Christy criteria for AOSD) or the PRINTO criteria. Similarly, we decided to have a tolerance for retaining studies that used classification criteria, but in which the final diagnosis was from the physician.

The protocol for SR3 was registered in PROSPERO (CRD42024534021, https://www.crd.york.ac.uk/prospero).

Data availability statement

The data extracted from included studies is available upon reasonable request.