Clinical, biological, and radiological changes under therapy, and criteria for clinical response and remission in adult-onset Still's disease: a systematic literature review informing the development of the European Alliance of Associations for Rheumatology (EULAR) criteria for assessing disease activity

Abstract

Objectives

A task force commissioned by the European Alliance of Associations for Rheumatology was set up for the Development And Validation of a composite disease actIvity score in adult-onset Still’s Disease (AOSD) (CLI113). A systematic literature review was undertaken to extract evidence regarding the disease activity assessment in AOSD and to identify potential candidate variables to be included in the criteria for the assessment of disease activity.

Methods

The PubMed MEDLINE, EMBASE, and Web of Science databases were screened for eligible articles published between January 1990 and September 2021. Abstracts from 2018 to 2021 EULAR, ACR, and ISSAID conferences were also browsed. We included randomised controlled trials (RCTs), quasi-RCTs, cohort studies and case series of ≥ 4 patients describing changes under therapy, with a minimum follow-up of 4 weeks. The risk of bias was assessed using the National Institute of Health Quality Assessment Tool. Results were synthesised descriptively.

Results

Sixty-three studies (including 2889 patients) were selected. Four were RCTs or quasi-RCTs, and 59 were observational studies. Most of the studies were retrospective (n = 58) and of poor quality. The most often reported clinical characteristics were fever, cutaneous rash, arthralgia and/or arthritis, hepatosplenomegaly, and sore throat. The most often reported laboratory parameters were leukocyte count, erythrocyte sedimentation rate, serum C-reactive protein and ferritin levels, all of which decreased with treatment. Definitions of response, remission and relapse, disease activity composite indices and patient-reported outcomes were heterogeneous.

Conclusions

Based on the identified variables, we can aim for a subsequent establishment and validation of a new measure to improve and standardise the management of patients with AOSD.

INTRODUCTION

Adult-onset Still’s disease (AOSD) is a rare systemic, inflammatory disorder of unknown aetiology affecting mainly young adults [1]. AOSD is characterised by daily high spiking fever, arthralgia or arthritis, evanescent salmon-pink rash and multiorgan involvement [1,2]. Laboratory findings usually show increased inflammatory markers and serum ferritin levels that are usually higher than those observed in other autoimmune, inflammatory, infectious or neoplastic diseases [3]. The course of AOSD can be diverse, and different patterns have been described so far as follows: (i) monocyclic pattern, characterised by a single systemic episode (30% of cases); (ii) polycyclic pattern, characterised by multiple, < 1 year lasting, flares, alternating with remissions (30%); and (iii) chronic pattern, related to a persistently active disease with chronic polyarthritis (40%) [[1], [2], [3], [4]]. However, these patterns are mainly based on case series and not on robust epidemiological studies [[5], [6], [7], [8], [9]]. Moreover, patients with AOSD may experience several severe life-threatening complications, such as macrophage activation syndrome (MAS), which are associated with a high mortality rate [10,11].

The treatment of AOSD remains largely empirical, due to the lack of controlled clinical trials. High doses glucocorticoids are usually the first line therapy when systemic symptoms predominate [12]. Despite this treatment, several patients experience recurrent flares with an evolution toward a chronic disease course and AOSD-related complications [3]. In this context, several studies reported the efficacy of biological disease modifying anti-rheumatic drugs (DMARDs) targeting interleukin (IL)-1 or IL-6 pathways in patients with refractory AOSD [[12], [13], [14], [15], [16], [17], [18], [19]]. Recently, a clinical trial, assessing primarily the safety, but also some efficacy endpoints of tadekinig alfa—a recombinant human IL-18 binding protein—was published [20]. However, due to the lack of validated clinical tools, an agreement is still missing concerning disease activity assessment, definition of refractory patients and disease remission.

A task force has been commissioned by the European Alliance of Associations for Rheumatology (EULAR) to develop the criteria for the assessment of disease activity in AOSD, a measure to be used in trials as well as in usual care, to ultimately improve and standardize the management of patients with AOSD [21]. This systematic literature review (SLR) was undertaken to inform the task force of evidence regarding the disease activity assessment in AOSD and potential candidate variables that could be included in the future criteria for the assessment of disease activity. To this end, we aimed to identify how clinical, radiological and biological features as well as disease activity and remission were assessed in clinical trials or observational studies in AOSD.

METHODS

Search strategy and studies selection

An SLR was performed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology [22] and in accordance with up-to-date EULAR standard operating procedures for EULAR-endorsed recommendations [23]. The search was carried out by 3 fellow authors (CGG, SM, PR), with guidance from the task force convenors (RG, EF), 1 methodologist (TS) and 2 task force lead investigators (BF, CG), using the PubMed, MEDLINE, EMBASE, and Web of Science databases. It was started in July 2018, with regular updates until September 2021 (the chosen time frame was from January 01, 1990, to September 2, 2021. Additionally, conference abstracts of the EULAR, American College of Rheumatology (ACR) and International Society of Systemic Auto-Inflammatory Diseases (ISSAID) annual conferences of the previous 4 years (2018-2021), and reference lists of included studies were screened. The relevant studies were selected first on titles and abstracts, then on full texts. The detailed search strategy and keywords used can be found in Supplementary Table S1. Eligible studies were randomised controlled trials (RCTs), quasi-RCT (ie trials in which treatment allocation was made by alternation, use of alternate medical records, date of birth or other expected methods), observational cohort studies and case series recruiting ≥ 4 patients with a minimum follow-up of 4 weeks. Only full-text versions published in the English language were retrieved. Preclinical studies were excluded.

Rather than using a population, intervention, control, and outcomes (PICO) statement [24] scheme in its narrowest sense to define the search questions, we formulated our question in a narrative manner (which features were used for outcome measurement in AOSD?). However, we applied the PICO scheme to define the following inclusion and exclusion criteria: (i) the target population was patients with AOSD; (ii) the intervention was the description of therapies used in AOSD; and (iii) the comparator could be placebo, glucocorticoids, conventional synthetic DMARDs (csDMARDs), biological DMARDS (bDMARDs) or the same therapy in different doses, formulations, regimens or treatment duration. However, as we were interested in outcomes and potential candidate variables for the new criteria for the assessment of disease activity, studies without a comparator were also included. (iv) The outcome domains included clinical, biological or radiological characteristics, and their possible change after therapy, as well as composite indices used to define clinical response or remission, and patient-reported outcomes (PROs).

Data extraction

The data extraction was performed by three fellow authors (CGG, SM, PR). Each parameter was extracted doubly and independently for assessment of study quality and evidence synthesis, using a predefined form. Any disagreement was resolved through discussion among the fellows and supervisors.

The following parameters were extracted: socio-demographic characteristics, clinical and radiological features, laboratory parameters, PROs, therapies (previous, current and cumulative corticosteroid/csDMARD/bDMARD exposure; drug tested; comparator), rates of remission and clinical response, and treatment discontinuation and the reason for it (intolerance, adverse events, lack of efficacy, persistent remission). For each outcome, pretreatment and posttreatment data were extracted, if applicable, and different time points were assessed when available: closest to 3, 6 and 12 months after treatment onset.

Data analysis

Quality was assessed through the Quality Assessment Tool proposed by the National Heart, Lung, and Blood Institute - US Department of Health & Human Services [25]. The level of evidence was assessed according to the recommendations of the Oxford Centre for Evidence-Based Medicine [26].

Information on candidate variables was synthesised descriptively. Owing to the vast heterogeneity of the included studies and the overall poor quality, a meta-analysis was not considered appropriate. Whenever there were several time points reported (closest to 3, 6 or 12 months), only the last follow-up data were considered for analysis of clinical and radiological variables. The best value (ie the strongest variation from baseline) during the course, independently of the time point, was chosen for the analysis of each laboratory parameter in each study. The disease activity scores and criteria of remission, already used in managing patients with AOSD, were extracted if prespecified at the beginning of the study and considering the results at the last follow-up.

Funding

This SLR was the first stage of the Development And Validation of a disease actIvity score in adult-onset Still’s Disease (DAVID) project funded by EULAR (project number CLI113).

Results

Systematic literature search output

The search identified 2895 abstracts. After duplicate exclusion, 2153 abstracts were screened, based on title and abstract. Hand search through references screening yielded 4 additional articles. From this first step, 164 articles were identified as potentially relevant and selected for full-text assessment. Finally, 63 selected articles were considered suitable for inclusion in the analysis (Fig 1 and Supplementary Table S2).

Fig 1.

Flow chart summary of the systematic literature review, article identification, screening and final selection. Footnote: *Among these 63 articles selected, 6 were retrieved, although there might potentially have been an overlap between different cohorts: 2 papers dealt with the same cohort but with different outcomes (Colafrancesco et al 2017 and Vitale et al 2019), and in 2 other cases, 2 articles from the same group of authors seemed to deal each time with the same cohort, but with an implemented number of patients in the most recent article (Ortiz-Sanjuan et al 2014 and Ortiz-Sanjuan et al 2015; Vignes et al 1998 and Vignes et al 2000, see Supplementary Table S2).

Studies characteristics

Out of the 63 included studies, 2 were RCTs, 2 were quasi-RCTs and 59 were observational studies (45 observational cohort studies and 14 case series). The majority of the articles were retrospective (n=58), whereas only 5 studies were prospective. Hence, the overall quality was poor, mainly related to their retrospective and observational design: only 2 papers were assessed as good and 3 additional ones as fair (Supplementary Table S2). This corresponds to a 2b level of evidence according to the Oxford Centre for Evidence-based Medicine.

Description of the study type, the country of origin, the number of patients, the gender distribution, the mean age and disease duration, the predominant disease pattern (systemic/articular), the therapy evaluated and its potential comparator(s), and quality assessment for each study are detailed in Supplementary Table S2.

Clinical and radiological variable outputs

Not all selected studies reported in detail clinical and radiological variables. The main AOSD manifestations assessed were fever, cutaneous rash, arthralgia and/or arthritis, hepatosplenomegaly, sore throat, lymphadenopathies (all items included in AOSD classification criteria), serositis and myalgia (Table 1) [[27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48]]. Pneumonitis, myocarditis, MAS and other complications were scarcely reported.

Table 1.

Main clinical and radiological variables

Main clinical and radiological variables		Definitions	Number of studies with available data:  • at baseline  • after intervention (ie treatment)	Number of patients with available data at baseline and after intervention	Study design (n studies with available data at baseline and after intervention)	Mean frequency of variable (%)		References
						At baseline	After intervention 0 (last follow-up visit)
Fever		• Not defined: n=37 • Body temperature > 39.5°C: n=1 ≥ 39°C: n=17 ≥ 38°C: n=3 AND/OR • Spiking: n=7 AND/OR • Duration ≥ 1 week: n=5 ≥2 weeks: n=4	• At baseline: n=58 • After intervention: n=21	478	• RCT or Quasi-RCT: n=3 • Observational studiesƗ: n=18	• 73.8 (9-100)	• 5.8 (0-25)	[20,27,28] [[29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46]]
Cutaneous rash		• Not defined: n=28 • Evanescent, salmon-pink, maculopapular or popular, predominating on the trunk and proximal limbs: n=4 OR • Meeting at least one of these features: n=10 OR • ‘Typical’, without precision: n=13 OR • Either “typical” or not: n=2	• At baseline: n=56 • After intervention: n=19	452	• RCT or Quasi-RCT: n=3 • Observational studiesƗ: n=16	• 60.8 (22-100)	• 6.7 (0-27)	[20,27,28] [[29], [30], [31], [32],[34], [35], [36], [37], [38],[40], [41], [42], [43], [44], [45], [46]]
Joint involvement	Arthritis	• Not defined: n=43 • Extent Mono, oligo or polyarticular: n=3 Polyarticular: n=4 AND/OR • Duration >2 weeks: n=2 Chronic: n=1	• At baseline: n=42 • After intervention: n=13	381	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=13	• 74.2 (33-100)	• 8.5 (0-42)	[[29], [30], [31], [32],35,37,38,[43], [44], [45], [46], [47], [48]]
	Arthralgia	• Not defined: n=41 • Polyarticular: n=2 • >2 weeks: n=3	• At baseline: n=34 • After intervention: n=10	354	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=10	• 79.6 (17-100)	• 10.1 (0-33)	[29,31,32,38,[43], [44], [45], [46], [47], [48]]
Sore throat		• Not defined: n=41 • Pharyngitis: n=2	• At baseline: n=43 • After intervention: n=6	217	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=6	• 45.5 (25-75)	• 0.3 (0-2)	[29,31,35,37,44,46]
Hepatomegaly		• Not defined: n = 22 • Confirmed by US or CT: n = 3 US, CT and/or MRI: n = 2 OR • Mild to moderate on US: n = 1 OR • Palpable or enlarged on US or CT: n=2	• At baseline: n=31 • After intervention: n = 7	307	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=7	• 33.5 (9-75)	• 4.4 (0-25)	[29,31,37,40,41,44,46]
Splenomegaly		• Not defined: n=32 • Confirmed by US or CT: n=3 OR • Palpable or detected by US or CT: n=1 OR • Slight to moderate, without precision: n=1	• At baseline: n=35 • After intervention: n=5	122	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=5	• 33.2 (6-75)	• 0.4 (0-2)	[31,37,41,44,46]
Lymphadenopathies		• Not defined: n=37 • Size >1 cm: n=1 Enlarged, without precision: n=2 AND/OR • Number >3: n=1 In ≥2 sites: n=4 AND/OR • Confirmed bs US and/or CT: n=2 AND/OR • Captured on physical examination: n=1	• At baseline: n=41 • After intervention: n=6	296	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=5	• 33.9 (11-51)	• 0.7 (0-2.5)	[28] [29,31,35,40,41]
Pleuritis		• Not defined: n=17 • Radiologically diagnosed by Chest X-ray: n= 1 CT: n=1 Chest X-ray and/or CT: n=4 OR • Diagnosed on clinical and/or radiological findings with Pleuritic pain and pleural effusion: n=3 Pleuritic pain, pleural rub or radiographic pleural effusion: n=2 Physical signs, without precision: n=1	• At baseline: n=29 • After intervention: n=4	205	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=3	• 17.3 (3-44.5)	• 0.3 (0-1.5)	[28] [29,35,40]
Pericarditis		• Not defined: n=16 • Radiologically diagnosed as pericardial effusion by Echocardiogram: n=3 Echocardiogram or CT: n=1 OR • Chest pain, pericardial rub or US pericardial effusion: n=5 OR • Captured on physical examination, without precision: n=1	• At baseline: n=26 • After intervention: n=5	209	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=4	• 20.2 (0-50)	• 0 (0-0)	[28] [29,35,37,40]
Myalgia		• Not defined: n=23	• At baseline: n=23 • After intervention: n=3	157	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=3	• 75.7 (50-100)	• 2.3 (0-9)	[29,35,44]

Frequencies are presented as mean percentages, with minimal and maximal values between brackets.

Whenever there were several time points reported (closest to 3, 6 or 12 months), only the last follow-up data were considered for analysis of clinical and morphological variables.

^ƗObservational studies included observational cohort studies and case series.

Noteworthy, only a few papers gave precise definitions of each of the reported manifestations, and definitions varied substantially from one study to another (Table 1). Moreover, the evolution of clinical and radiological variables over time, especially after treatment of AOSD, was reported in a minority of studies, including predominantly retrospective observational studies. As an example, fever, which was the most frequently reported clinical manifestation at baseline (58 studies), was assessed after treatment in only 21 papers, including 2 RCTs and 1 quasi-RCT (Table 1).

In studies assessing outcomes before and after treatment, fever, cutaneous rash, arthralgia/arthritis and myalgia appeared to be the most common manifestations at baseline and their frequency decreased dramatically after treatment, especially fever (Table 1). Less frequently reported manifestations including hepatomegaly, splenomegaly, lymphadenopathies, pleuritis and pericarditis were less prevalent at baseline but seemed to resolve in a large majority of patients after treatment (Table 1).

Among serositis, peritonitis was reported in only 1 study [49]. Another 1 assessed ascites [50], but none of these 2 studies provided data after treatment. Abdominal pain was reported as a clinical manifestation of AOSD in 7 additional studies [3,[50], [51], [52], [53], [54], [55], [56]]. Pneumonitis and myocarditis were respectively reported in 5 [29,[56], [57], [58], [59], [60]] and 3 articles [3,60,61], with a mean frequency at baseline of 18 and 17%, without available follow-up data. Weight loss was reported in 38.6% of patients at baseline in 2 papers, without follow-up data [50,52]. MAS was present in 104 patients and disseminated intravascular coagulation in 45 patients as described in 16 [3,[29], [30], [31], [32],[57], [58], [59], [60], [61], [62], [63], [64], [65], [66]] and 7 articles [57,59,60,63,[66], [67], [68]], respectively. Other more seldom complications including multiorgan failure, acute respiratory distress syndrome and fulminant hepatitis were reported in only a couple of papers [60,66].

Laboratory parameters outputs

Of the studies reporting laboratory parameters, all evaluated biological responses to treatment, except one, which evaluated biomarkers for disease flare (relapse) in patients treated with tocilizumab [69].

‘Routine’ laboratory parameters were the most often reported, especially leukocyte number, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) and ferritin level (Table 2) [20,27,[29], [30], [31], [32], [33], [34], [35], [36], [37],[39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49],57,58,61,66,68,[70], [71], [72], [73], [74], [75], [76], [77]]. Other biochemical parameters, such as the number or percentage of polymorphonuclears, and the level of haemoglobin, platelets or transaminases, were less often reported. All ‘routine’ laboratory parameters decreased with treatment and clinical response, except haemoglobin which increased. Furthermore, the study by Yamada et al [69], which evaluated biomarkers for disease flare (relapse) in patients treated with tocilizumab, suggested that this therapy might lower the values of some biological markers such as ESR, CRP or ferritin. Indeed, these biomarkers were significantly more elevated in patients treated without tocilizumab, whereas their changes were modest in patients receiving tocilizumab. In contrast, the white blood cells (WBC) and lactate dehydrogenase (LDH) levels were similarly elevated in both groups.

Table 2.

Routine and/or classical laboratory parameters

Laboratory parameters	Definition, operationalization, measure	Number of studies with available data: • At baseline • After intervention (ie treatment)	Total number of patients┼	Study design (n studies)	Kinetic	Statistically significant change, if reported	References
Leukocytes	• Mean number (SD) /mm3 or × 109/L OR Median number (range) in 1 study • Measured in plasma	• At baseline: n=41 • After intervention: n=18	346	• RCT or Quasi-RCT: n=2 • Observational studiesƗ: n=16	Decrease with treatment and clinical response	Statistically significance reported:  • < 2-fold decrease: 6studies  • > 2-fold decrease: 2 studies No statistics done, but meaningful:  • < 2-fold decrease: 4 studies  • > 2-fold decrease: 6 studies	[20,32,41,42,58,70] [45,71] [30,31,34,72] [27,[42], [43], [44],57,73]
Polymorphonuclears	• Mean percentage (%) of total number of leukocytes OR Mean number (SD) x 109/L • Measured in plasma	• At baseline: n=12 • After intervention: n=4	69	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=3	Decrease with treatment and clinical response	Statistically significance reported:  • < 20% decrease: 2 studies No statistics done, but meaningful:  • > 20% decrease: 2 studies	[20,30] [43,44]
ESR	• Mean (SD or range) rate in mm at 1st hour • Measured in plasma	• At baseline: n=45 • After intervention: n=24	534	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=24	Decrease with treatment and clinical response	Statistically significance reported:  • Normalisation with a > 3-fold decrease in 8 studies  • Decrease without complete normalisation in one study No statistics done, but meaningful:  • Normalisation in 10 studies  • No normalisation in 2 studies (but the time point was 12 weeks in one of these studies)  • With a > 2-fold decrease in 2 studies, and a > 3-fold decrease in 6 studies Not statistically significant, but meaningful:  • Normalisation in 1 study (> 2-fold decrease)  • No normalisation in 1 study (< 2-fold decrease)  • Normalisation in most patients in 1 study	[32,[39], [40], [41],58,70,71,74] [30] [29,33,34,43,44,49,61,73,75] [27,31] [45] [36] [35]
CRP	• Mean level (SD) in mg/L OR Median (range) in mg/L • Measured in plasma	• At baseline: n=46 • After intervention: n=24	577 (603)*	• RCT or Quasi-RCT: n=2 • Observational studiesƗ : n=22	Decrease with treatment and clinical response	Statistically significance reported in 7 studies, with a normalisation in all:  • < 10-fold decrease in 1 study  • ≥ 10-fold decrease in 3 studies  • ≥ 20-fold decrease in 3 studies  • Data not exploitable in 2 studies (significant change, but figure with no scale) No statistics done, but meaningful in 12 studies:  • Normalisation in 8 studies  • No normalisation but decrease in 5studies  • With a < 10-fold decrease in 4 studies, a ≥ 10-fold decrease in 5 studies, and a ≥ 20-fold decrease in 2 studies Not statistically significant, but meaningful in 1 study  • No normalisation but decrease (normalisation in 25-50% at week 12), with a < 10-fold decrease (but data are in median) in Gabay et al 2018  • Normalisation in most patients in 1 study	[58] [41,70,74] [32,39,40] [27,30] [33,34,37,[42], [43], [44],68,73] [29,31,47,71,76] [20] [35]
Ferritin	• Mean level (SD) in ng/mL OR Median (range) in mg/L • Measured in plasma	• At baseline: n =47 • After intervention: n=22	621 (647)⁎⁎	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=21	Decrease with treatment and clinical response	Statistically significance reported in 7 studies:  • Normalisation in 6 studies  • No normalisation but decrease in 1 study (normalisation in 38% at W12)  • With a < 5-fold decrease in 1 study, a ≥ 5-fold decrease in 2 studies, and a ≥ 10-fold decrease in 4 studies No statistics done, but meaningful in 10 studies  • Normalisation in 10 studies  • Normalisation in most patients in 1 study  • No normalisation but decrease in 2 studies  • With a < 5-fold decrease in 2 studies (in the chronic articular group only for Colina et al 2011), a ≥ 5-fold decrease in 3 studies and a≥ 10-fold decrease in 3 studies (of which the systemic group for Colina et al 2011) Not statistically significant, but meaningful in 2 studies  • With normalisation and a < 10-fold decrease  • With normalisation and a ≥ 10-fold decrease Not exploitable in 1 study (figure with no scale)	[29,32,40,41,68,70] [20] [30,31,33,34,37,43,44,47,57,77] [35] [47,49] [58] [45] [27]
Glycosylated ferritin	• Mean (SD) % of ferritin that is glycosylated • Measured in plasma	• At baseline: n =3 • After intervention: n=3	78	• RCT or Quasi-RCT: n=0 • Observational studiesƗ : n=3	Can remain low or normalize in patients in remission	Remained low in patients in remission in 2 studies In the 3rd study, during the remission of 5 patients the GF normalized in 4 patients (up to 31%; 40%; 45%, and 50%) but remained low at 19% in the fifth patient	[42,77] [66]
Haemoglobin	• Mean level (SD) in g/dL • Measured in plasma	• At baseline: n =17 • After intervention: n=7	194	• RCT or Quasi-RCT: n=0 • Observational studiesƗ : n=7	Increase with treatment and clinical response	Statistically significance reported in 2 studies:  • <2 g/dL increase in both No statistics done, but meaningful in 4 studies:  • <2 g/dL increase in 3 studies  • >2 g/dL increase in 1 study Not exploitable in 1 study (figure without a scale)	[40,41] [31,33,44] [43] [27]
Platelets	• Mean number (SD) in G/L or .103/mm3 • Measured in plasma	• At baseline: n =9 • After intervention: n=3	78 (104)⁎⁎⁎	• RCT or Quasi-RCT: n=0 • Observational studiesƗ : n=3	Decrease with treatment and clinical response	No statistics done, with a < 2-fold decrease in 2 studies Data not exploitable (figure without a scale) in 1 study	[31,43] [27]
Transaminases	• X-fold the upper limit of normal • Or mean (SD), UI/L • Or elevation in number (%) of total number of patients • Measured in plasma	• At baseline: n=22 • After intervention: n=5	195	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=4	Normalize in 92-100% of patients in remission	Giampietro 2013: normalised in 92.9% of patients; no statistics done Colafrancesco 2017: normalised in 95% of patients at 3 months, 93.6% of patients at 6 months, 97% of patients at 12 months; no statistics done Gabay 2018: normalized in the 2 patients (out of 23) with initially elevated AST (elevated > 3-fold upper normal limits in 2 patients); no statistics done Fitzgerald 2015: normalised in 3 patients (out 4), no data for the last; no statistics done Laskari 2020: normalised in all patients at last visit after 12 months	[48] [29] [20] [44] [31]

Results are expressed as X-fold increase (or decrease) or X% increase (or decrease). Owing to the heterogeneity of the time points (closest to 3, 6 or 12 months), the study types and the laboratory reference values among the studies, the best value (ie the strongest variation from baseline) during the course, independently of the time point, was chosen for each laboratory parameter in each study.

^┼Total number of patients from the studies for which data after intervention were available.

^ƗObservational studies included observational cohort studies and case series. IQR, interquartile range; SD, standard deviation.

^⁎There were 577 patients without including the patients of Kaneko et al 2019; otherwise, 603. A number of patients are given with and without including the patients of Kaneko 2019, because although given, the data were not directly exploitable.

^⁎⁎There were 621 patients without including the patients of Kaneko et al 2019; otherwise, 647. A number of patients are given with and without including the patients of Kaneko 2019, because although given, the data were not directly exploitable.

^⁎⁎⁎There were 78 patients without including the patients of Kaneko et al 2019; otherwise, 104. A number of patients are given with and without including the patients of Kaneko 2019, because although given, the data were not directly exploitable.

The levels of glycosylated ferritin were reported in 3 observational studies. They remained low in 2 studies but normalized with remission in a majority of patients in the third study.

Five studies reported cytokine levels at baseline and after response to treatment (Table 3). All decreased with treatment and clinical response, except IL-1Ra which increased immediately after treatment but seemed to decrease with time. Only 1 study reported alarmins such as S100 proteins or serum amyloid A protein (Table 3).

Table 3.

Nonroutine immunological laboratory parameters

Laboratory parameters	Definition, operationalization, measure	Number of studies with available data:  • At baseline  • After intervention (ie treatment)	Total number of patients	Study design (n studies)	Kinetic	Statistically significant change, if reported	References
Cytokines
IL-1β	• Mean (SD) or median (IQR) level, in pg/mL • Measured in plasma	• At baseline: n =2 • After intervention: n=1	4	• RCT or Quasi-RCT: n=0 • Observational studiesƗ: n=1	Decrease with treatment and clinical response	IL-1 was elevated only in patients 2 and 3 and normalized with treatment	[73]
IL-1Ra	• Mean (SD) or median (IQR) level, in pg/mL • Measured in plasma	• At baseline: n =3 • After intervention: n=3	33	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=2	Increase (immediately) after treatment, but seems to decrease with time	Gabay 2018: moderate and nonstatistically decrease from baseline at week 12 (median (IQR): 40.2 (25.8; 50.6) versus 35.1 (27.7; 48.4), p = 0.61) Kötter 2007 (4 patient case series): IL-1RA increased after the administration of anakinra (which is an IL-1 receptor antagonist and can be measured in the serum) but was within the normal range before the drug was introduced in 2 of the 3 patients analysed. It was elevated in patient 3 before treatment but increased further with anakinra. Fitzgerald 2005: decrease 11 days after anakinra was stopped and increased again 5 days after anakinra was restarted (relapse after anakinra’s discontinuation)	[20] [73] [44]
IL-6	• Mean (SD) or median (IQR) level, in pg/mL • Measured in plasma	• At baseline: n =4 • After intervention: n=4	37	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=3	Decrease with treatment and clinical response	Gabay 2018: the levels of IL-6 significantly decreased at week 12, as compared with baseline levels (median (IQR): 4.7 (3.1; 9.8) versus 2.3 (1; 4.5), p = 0.007) Saviola 2010: IL-6 levels (evaluated in one patient) diminished to the normal level after 2 months of the treatment with clarithromycin (baseline value 39.5 pg/ml (normal values < 12.5 pg/ml)). Kötter 2007(4 patient case-series): IL-6 was elevated only in patient 3 before anakinra and normalized with therapy. Fitzgerald 2005: IL-6 decreased with anakinra treatment, was elevated during relapse after treatment discontinuation, and decreased again after reintroduction of treatment	[20] [75] [73] [44]
IL-18	• Mean (SD) or median (IQR) level, in pg/mL • Measured in plasma	• At baseline: n =3 • After intervention: n=3	33	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=2	Decrease with treatment and clinical response	Gabay 2018: serum levels of free IL-18 were detected in seven patients at baseline. Among these patients, four exhibited a clinical response. Free IL-18:  - was undetectable in all these patients at the final blood assessment  - remained elevated in two of three patients who failed to respond to tadekinig alfa. Kötter 2007 (4 patient case series): very high in all patients when their disease was active (before the initiation of anakinra) and returned to significantly lower levels when disease remission was achieved Fitzgerald 2005: data available for 1 patient only; elevated levels at baseline, which decreased under treatment with anakinra	[20] [73] [44]
TNF-α	• Mean (SD) or median (IQR) level, in pg/mL • Measured in plasma	• At baseline: n =3 • After intervention: n=3	34	• 1 RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=2	Decrease with treatment and clinical response	Gabay 2018: the levels of TNF-α decreased at week 12, as compared with baseline levels, but the difference was not statistically significant (median (IQR): 3 (2.6; 3.6) vs 2.7 (2; 3.4), p = 0.09) Saviola 2010: levels diminished to the normal level after 2 months of the treatment with clarithromycin Kötter 2007 (4 patient case series): TNF-α was slightly elevated in all patients before anakinra treatment and later normalized (no data on parametric statistics)	[20] [75] [73]
Other laboratory parameters
S100A12 protein	• Median (IQR) level in pg/mL • Measured in plasma	• At baseline: n =1 • After intervention: n=1	23	• 1 RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=0	Levels were significantly decreased at week 12, as compared with baseline levels	Levels were significantly decreased at week 12, as compared with baseline levels (median (IQR): 122.8 (51.1; 220) versus 67.5 (32.1; 125.9) ng/mL, p = 0.01)	[20]
S100A8/A9 protein	• Median (IQR) level in pg/mL • Measured in plasma	• At baseline: n =1 • After intervention: n=1	23	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=0	Levels were significantly decreased at week 12, as compared with baseline levels	Levels were significantly decreased at week 12, as compared with baseline levels (median (IQR): 6621 (3271; 11,436) versus 4,603 (2576; 6018) ng/mL, p = 0.01)	[20]
SAA protein	• Median (IQR) level in pg/mL • Measured in plasma	• At baseline: n =1 • After intervention: n=1	23	• RCT or Quasi-RCT: n=1 • Observational studiesƗ: n=0	Levels were significantly decreased at week 12, as compared with baseline levels	Levels of were significantly decreased at week 12, as compared with baseline levels (median (IQR): 235.4 (56.6; 452.1) versus 52.4 (10.4; 240.8) µg/mL, p = 0.09)	[20]

IQR, interquartile range; SAA, serum amyloid A; SD, standard deviation.

^ƗObservational studies included observational cohort studies and case series.

Results are expressed as X-fold increase (or decrease) or X % increase (or decrease). Owing to the heterogeneity of the time points (closest to 3, 6 or 12 months), the study types and the laboratory reference values among the studies, the best value (ie the strongest variation from baseline) during the course, independently of the time point, was chosen for each laboratory parameter in each study.

Response criteria used in AOSD studies

Many criteria of response have been used in AOSD so far, reflecting the heterogeneity of the disease with different clinical tools assessing either systemic features or joint involvement, as reported in Table 4. Clinical trials developed their own criteria of response, mirroring the lack of a validated disease activity score [20,27]. Finally, different PROs were used in only a few studies.

Table 4.

Response criteria used in AOSD studies.

Criteria of response already used in AOSD	Definition of response criteria	Number of studies with available data:  • At baseline  • After intervention (ie treatment)	Total number of patients	Study design (n studies)	Statistically significant change, if reported	References
Definition of response in RCT of tadekinig alfa	Early predicted criteria of response at 3 weeks were normalisation of body temperature and decrease by 50% of the baseline CRP levels or normalisation of CRP values to <5 mg/L. Response to therapy at 12 weeks was predefined as an improvement of joint count (both Swollen Joint Count and Tender Joint Count according to a 44-joint assessment) by ≥20% from baseline values, and a 70% decrease in CRP levels compared with baseline values (or reduction to normal levels) or normalisation of ferritin.	• At baseline: n=1 • After intervention: n=1	23	• Quasi-RCT: n=1	No statistics analysis performed, but meaningful: At week 3, 5 of 10 patients receiving 80 mg and 6 of 12 patients receiving 160 mg achieved the predefined response criteria.	[20]
Definition of response in RCT of tocilizumab	The ACR core (ACR20, 50, 70) combined with systemic feature score. Systemic feature score consisted of 5 clinical (fever, rash, lymphadenopathy, hepatosplenomegaly and serositis) and 5 laboratory assessments (ESR, CRP; leucocytes count, haemoglobin and platelet). Each clinical feature was assigned a score of 1 (present) or 0 (absent).	• At baseline: n=1 • After intervention: n=1	27	• RCT: n = 1	Statistics analysis performed: ACR50 response at week 4 was achieved in 61.5% in the tocilizumab group and 30.8% in the placebo group. Systemic feature score reduced at week 12 were –4.1 in the tocilizumab group and –2.3 in the placebo group.	[27]
Definition of response in RCT of canakinumab	Change in disease activity score (ΔDAS28(ESR)>1.2) at week 12 and a modified adapted ACR variable (requiring ACR 30% response (ACR30) in addition to no intermittent fever in the preceding week and no more than one out of 7 variables worsening.	• At baseline: n=1 • After intervention: n=1	36	• RCT: n = 1	Statistics analysis performed: After 12 weeks, a higher percentage of patients treated with canakinumab showed a reduction of the DAS28(ESR) of more than 1.2 at week 12 (66.7% vs 41.2%, respectively).	[28]
Systemic score* and systemic score modified according of Rau et al⁎⁎	Systemic score: Fever, typical rash, pleuritis, pneumonia, pericarditis, hepatomegaly or abnormal liver function tests, splenomegaly, lymphadenopathy, leukocytosis > 15,000/mm3, sore throat, myalgia, and abdominal pain. Each clinical feature is assigned a score of 1 (present) or 0 (absent). Modified systemic score: Fever, evanescent rashes, sore throat, arthritis, myalgia, pleuritis, pericarditis, pneumonitis, lymphadenopathy, hepatomegaly or abnormal liver function tests, elevated leukocyte count > 15,000/µl, and serum ferritin > 3000 µg/L. Each clinical feature is assigned a score of 1 (present) or 0 (absent).	• At baseline: n = 8 • After intervention: n = 6	366	• Observational studiesƗ: n=8	Statistics analysis performed: > 2-fold decrease: 6 studies (Iliou, 2013, Colafrancesco 2017; Song 2017, Hu 2020, Vitale 2020) No statistics, but meaningful: >2-fold decrease in 1 study (Cavalli 2019)	[29,30,35,37,54,74,78,79]
DAS28	DAS28 calculated as 0.56*sqrt(t28) + 0.28*sqrt(sw28) + 0.70*Ln(ESR) + 0.014*GH The criteria of response combined the reduction of DAS28 with a resolution of fever as an outcome to be assessed (Puechal 2011; Cipriani 2014).	• At baseline: n = 5 • After intervention: n = 5	210	• Observational studiesƗ: n = 4 • RCT: n = 1	Statistics analysis performed: > 2-fold decrease in 2 studies (Colafrancesco 2017; Cipriani 2014, Vitale 2020) Statistics analysis performed: < 2-fold decrease in 1 study (Kedor 2020) No statistics, but meaningful: < 2-fold decrease in 1 study (Puéchal 2011)	[29,33,35] [28] [34]
ACR criteria	ACR20, ACR50 and ACR70 responses defined as an improvement of 20%, 50% and 70% in the number of tender and swollen joints and 20%, 50% and 70% improvement in at least 3 of the remaining 5 variables of the ACR core set, respectively. Lequerre 2007: Response was defined as a resolution of systemic symptoms and an improvement of the ACR score by at least 20% in patients with AOSD. The response was defined as a ‘partial response’ if the ACR score improvement was less than 50% and was defined as a ‘complete response’ if the ACR score improved by 50% or more.	• At baseline: n =4 • After intervention: n=4	69	• Observational studiesƗ: n=3 • RCT: n =1	No statistics, but meaningful:  A percentage of patients achieved ACR20, 50, and 70. Statistics analysis performed:  A percentage of patients achieved ACR20, 50 and 70 (Kedor 2020)	[36,71,75] [28]
Physicians global assessment of disease	Physician’s global assessment by using Visual Analogue Scale	• At baseline: n =4 • After intervention: n=4	101	• RCT or Quasi-RCT: n=3 • Observational studiesƗ: n=1	Statistics analysis performed: < 2-fold decrease in 3 studies (Kaneko 2018; Gabay 2018; Kedor 2020) No statistics, but meaningful: A percentage of patients achieved a 70% of improvement in 1 study (Lequerré 2007)	[20,27,28] [71]
Drug retention rate	The proportion of patients who maintain the same drug in a given time period	• At baseline: n =3 • After intervention: n=3	395	• Observational studiesƗ: n=3	Statistics analysis performed: A percentage of patients showed an overall good long-term drug retention rate.	[63,72,80]
SF36, physical and mental health domains	36-item patient-reported survey of patient health	• At baseline: n =2 • After intervention: n=2	58	• Quasi-RCT: n=1 • RCT: n=1	Statistics analysis performed: Decrease of SF 36, mainly in physical health domains.	[28,76]
Clinical response combining glucocorticoid dosage and laboratory markers	Reduction of glucocorticoid dosage and laboratory markers	• At baseline: n =1 • After intervention: n=1	39	Observational studiesƗ: n=1	Statistics analysis performed: Decrease of glucocorticoid dosage and laboratory markers.	[58]
Patient global assessment	patient’s global assessment using Visual Analogue Scale	• At baseline: n =3 • After intervention: n=3	65	• RCT or Quasi-RCT: n=2 • Observational studiesƗ: n=1	Statistics analysis performed: <2-fold decrease in 2 studies (Kaneko 2018; Gabay 2018) No statistics, but meaningful: A percentage of patients achieved a 70% of improvement in 1 study (Lequerré 2007)	[20,27] [71]
Patient assessment of pain	patient’s pain assessment using Visual Analogue Scale	• At baseline: n =4 • After intervention: n=4	101	• RCT or Quasi-RCT: n=3 • Observational studiesƗ: n=1	Statistics analysis performed: < 2-fold decrease in 2 studies (Kaneko 2018; Gabay 2018; Kedor 2020) No statistics, but meaningful: A percentage of patients achieved a 70% of improvement in 1 study (Lequerré 2007)	[20,27,28] [71]
Patient assessment of fatigue	Patient’s fatigue assessment using Visual Analogue Scale	• At baseline: n =1 • After intervention: n=1	23	• Quasi-RCT: n=1	Statistics analysis performed: < 2-fold decrease.	[20]
HAQ	Health assessment questionnaire to evaluate disability.	• At baseline: n =2 • After intervention: n=2	27	• RCT: n=2	Statistical analysis performed: The improvements in HAQ were not different between the groups	[27,28]

ƗObservational studies included observational cohort studies and case series.

^⁎Pouchot et al. Medicine (Baltimore). 1991;70:118-136.

^⁎⁎Rau et al. J Rheumatol. 2010;37:2369-2376.

Definitions of response in clinical trials

In a clinical trial assessing tadekinig alfa in AOSD, early predictive criteria of response were the normalisation of body temperature and the decrease by 50% of the baseline CRP levels or normalisation of CRP values to <5 mg/L, after 3 weeks. The response to therapy at week 12 was defined as an improvement of the joint count, considering both swollen joint count and tender joint count according to a 44-joint assessment, by ≥20% from baseline values and a 50% decrease in CRP levels compared with baseline values (or reduction to normal levels) or normalisation of ferritin [20].

In a clinical trial assessing tocilizumab in AOSD, response to treatment was evaluated by combining the ACR core set (ACR20, 50, 70) with a systemic feature score at week 12. The latter consisted of 5 clinical (fever, rash, lymphadenopathy, hepatosplenomegaly and serositis) and 5 laboratory assessments (ESR, CRP; WBC, haemoglobin and platelet). Each feature was assigned a score of 1 (present) or 0 (absent) [27].

A recent clinical trial investigated the efficacy of canakinumab in AOSD [28]. After 12 weeks, the primary outcome was a reduction of disease activity score in 28 joints (DAS28) of more than 1.2. Clinical response was also evaluated by EULAR and ACR response criteria developed for rheumatoid arthritis (RA) and a modified adapted ACR variable (requiring ACR30 response in addition to no intermittent fever in the preceding week and no more than 1 out of 7 variables worsening by more than 30%) [33].

RA-derived response criteria

Similar to what has just been described in clinical trials [27,28], DAS28 was used to assess the disease activity in some observational retrospective studies, alone or in combination with the evaluation of fever [29,[33], [34], [35]]. In these works, tocilizumab, anakinra and canakinumab induced a reduction of DAS28 associated with the disappearance of fever in the majority of patients.

Three further studies used the ACR core set to evaluate the response to therapies in AOSD [36,71,75]. In a prospective observational study, the clinical response to anakinra was defined as the resolution of systemic symptoms and an improvement of the ACR score by at least 20% [71]. The clinical response was deemed as a ‘partial response’ if the ACR score improvement was less than 50%, whereas a ‘complete response’ if the ACR score improved by 50% or more.

AOSD-specific scores

The systemic score was proposed to evaluate the systemic disease in AOSD [7], and its subsequent modification by Rau [81] was frequently used to evaluate disease activity and response to therapies [29,30,35,37,54,74,78,79]. In these studies, a significant reduction of the modified systemic score was observed during the follow-up [29,30,35,37,54,74]. In one publication, remission was defined as a modified systemic score = 0, partial response as a reduction < 2 or at least 20% of improvement and a good response as a decrease > 2 [74].

Physician’s global assessment of disease

In clinical trials assessing the efficacy of tadekinig alfa, tocilizumab and canakinumab [20,27,28], the physician’s global assessment of disease using the visual analogue scale (VAS) decreased significantly in patients treated with the study drug [20,27,28]. Similarly, in an observational prospective study, a percentage of patients treated with anakinra achieved a 70% of improvement in physician’s global assessment of disease [71].

Patient-reported outcomes

In a study assessing the efficacy of anakinra in AOSD, the short-form health survey (SF-36), a 36-item patient-reported survey of patient health, was used to assess the health status. After 24 weeks, anakinra induced an improvement in SF-36, mainly in the physical health domain [76]. Similarly, in the clinical trial of canakinumab, a reduction of physical and mental domains of SF-36 was reported [28]. Clinical trials of tadekinig alfa, tocilizumab and canakinumab in AOSD used VAS to assess pain and reported a significant reduction of this score under treatment [20,27,28]. Additionally, in an observational prospective study, a percentage of patients treated with anakinra achieved a 70% improvement in VAS pain [71]. Similarly, patient global assessment of disease using VAS improved during follow-up [20,27,71]. The patient’s fatigue assessment using VAS was also evaluated and was shown to decrease under tadekinig alfa therapy [20]. Finally, a reduction of the health assessment questionnaire (HAQ) value was recorded in clinical trials of tocilizumab and canakinumab [27,28].

Definition of clinical response combining glucocorticoid-sparing effect and laboratory markers

In a retrospective study, the effectiveness of tocilizumab was assessed by combining the reduction of laboratory markers and glucocorticoid dosage [58]. Patients treated with tocilizumab showed significantly reduced prednisolone uptake and decreased values of leucocytes, CRP, ESR and ferritin [58].

Drug retention rate

The drug retention rate (DRR) was assessed for evaluating the effectiveness of drugs in AOSD in two retrospective observational studies [72,80]. The cumulative DRR of anakinra was 44.6% and 49.4% after 60 months of follow-up [72,80]. In another study, the DRR of calcineurin inhibitors was estimated to be 71% after 5 years of follow-up [63].

Disease-related health states

Remission

Different definitions of remission were developed in managing patients with AOSD, as shown in Table 5. Several retrospective observational works proposed the definition of remission in AOSD, as the complete disappearance of clinical and laboratory signs of disease [29,31,38,[48], [49], [50], [51], [52], [53], [54],56,57,59,62,64,66,67,70,[82], [83], [84], [85], [86]].

Table 5.

Criteria of remission and relapse are already used in AOSD.

Criteria of remission already used in AOSD		Definition of remission criteria	Number of studies with available data:  • At baseline  • After intervention (ie treatment)	Total number of patients	Study design (n studies)	Statistically significant change, if reported	References
Remission considering clinical and biological features		Remission defined as the disappearance of clinical and laboratory signs of disease. Fautrel 2005: Remission was defined as a complete resolution of all clinical and biological AOSD-related symptoms, except joint erosion. Zhu 2009: Remission was defined as the absence of articular, systemic, and laboratory evidence of disease activity noted during chart review, for at least 2 consecutive months, regardless of the current therapy. Giampietro 2013: Complete remission defined as all clinical and biologic AOSD symptoms disappearing after treatment.	• At baseline: n=23 • After intervention: n=18	1187	Observational studiesƗ: n=23	No statistics, but meaningful: A percentage of patients, about 30%, achieved complete remission.	[29,31,38,[48], [49], [50], [51], [52], [53], [54],56,57,59,62,64,66,67,70,[82], [83], [84], [85], [86]]
Remission considering clinical and biological features, and medications		Remission defined as the disappearance of clinical and laboratory signs of disease, combined with a reduction of concomitant therapy Nordstrom 2012: Remission defined as patient afebrile (≤ 37°C body temperature, measured twice from armpit), in the absence of NSAIDs 24 h before measurement, decrease of CRP and ferritin to reference limits and normal swollen and tender joint counts. A full response was defined as body temperature ≤ 37°C, CRP ≤ 10 mg/l, and ferritin ≤ 200 μg/l female, ≤ 275 μg/l male, and normal swollen and tender joint counts. Cavalli 2015: Complete response was defined as the absence of articular and systemic manifestations of AOSD with normalization of inflammatory indexes (CRP and ESR), and with a reduction of the corticosteroid dose of at least 50% for at least 2 months.	• At baseline: n=2 • After intervention: n=2	42	• Quasi-RCT: n=1 • Observational studiesƗ: n=1	No statistics, but meaningful: A percentage of patients achieved complete remission.	[61,76]
Remission considering EULAR definitions of remission		EULAR remission defined as DAS28 < 2.6	• At baseline: n=4 • After intervention: n=4	76	• Observational studiesƗ: n=3 • RCT: n=1	No statistics, but meaningful: A percentage of patients achieved EULAR remission in 3 observational studies Statistics analysis performed: A significant percentage of patients achieved EULAR remission in RCT	[33,34,39] [28]
Long-term remission, achievement of monocyclic pattern*		The achievement and maintenance of remission during the follow-up. Colina 2011: Monocyclic pattern defined as the achievement of a symptom-free period, during the whole follow-up. Gerfaud-Valentin 2014: Controlled disease was defined as clinically asymptomatic AOSD with no biological inflammatory syndrome for at least 1-y follow-up. Ruscitti 2016: A monocyclic course was defined as a single episode for more than 2 months but less than 1 year followed by sustained remission, defined as the complete disappearance of systemic symptoms and normalization of laboratory evidence of disease activity through the entire follow-up period.	• At baseline: n =13 • After intervention: n=13	1118	• Observational studiesƗ: n=13	No statistics, but meaningful: A percentage of patients achieved the monocyclic pattern during the follow-up, maintaining a long-term remission.	[3,47,[49], [50], [51], [52],56,59,62,64,66,86,87]
Partial remission		Partial remission was defined as the achievement of not complete remission. Kong 2010: Partial remission was defined as partial remission of clinical symptoms, and more than 50% decrease in laboratory examinations such as leukocytosis, elevated liver enzymes, ESR, and ferritin. Giampietro 2013: Partial response defined as some improvement was noted by the physician in charge of the patient, but some systemic, articular, or biological features persisted after therapy started. Rossi-Semeraro 2015: Partial response was retained when a complete response was not achieved but the clinical improvement was evident according to the treating physician.	• At baseline: n =8 • After intervention: n=7	341	• Observational studiesƗ: n=8	A percentage of patients achieved partial remission	[29,31,[48], [49], [50],67,82,83]
Relapse considering clinical features		The reappearance of disease manifestation after achieving remission. Mitamura 2009: Recurrence was defined when at least 3/4 criteria appeared: body temperature higher than 37°C for over a week, arthralgia at more than 2 joints, erythematous rash, and serum CRP level higher than 1.0 mg/dl. Colina 2011: Active systemic disease required 2 or more of the following: temperature higher than 39°C; typical rash; serositis; leukocytosis 10,000/mm3 (with neutrophilia 90%); reticuloendothelial involvement; or ESR. Active articular disease requires 2 or more of the following: morning stiffness for 30 min, polyarthralgia, or synovial effusion. Liu 2015: Relapse was defined as patients with AOSD who had achieved complete remission and who redeveloped clinical symptoms and laboratory findings. Kalyoncu 2016: Relapse was defined as an active disease status according to the physician’s global assessment	• At baseline: n =5 • After intervention: n=5	557	• Observational studiesƗ: n=5	No statistics, but meaningful: A percentage of patients relapsed during the follow-up	[47,49,57,86,87]
Relapse considering clinical features and medications		The reappearance of disease manifestation after achieving the remission and the concomitant administration of therapies. Nishina 2015: Relapse was defined as a worsening in disease activity, which required an increase of at least 50% in the glucocorticoid dose or a restart of glucocorticoids when the patient was not taking glucocorticoids. Ruscitti 2016: Relapse was characterized by systemic flares occurring after remission, the need for any additional treatment and/or any increased dosage of drugs as a flare of the disease.	• At baseline: n =5 • After intervention: n=5	448	• Observational studiesƗ: n=5	No statistics, but meaningful: A percentage of patients relapsed during the follow-up Statistical analysis performed: a percentage of patients had a lower rate of event-free survival	[3,31,56,63,65]
Outcome in life-threatening manifestations		The clinical outcome in patients with AOSD who were admitted to the intensive care unit was described as leaving the intensive care unit without further treatment.	• At baseline: n =1 • After intervention: n=1	20	Observational studiesƗ: n=1	No statistics, but meaningful: A percentage of patients left the intensive care unit without further treatment.	[60]

ƗObservational studies included observational cohort studies and case series.

^⁎Cush JJ, et al. Arthritis Rheum 1987;30(2):186-194.(4).

In some studies, the remission was reported considering the resolution of some specific clinical and laboratory features, including fever, salmon pink rash and polyarthritis, considered to be the major and more relevant symptoms, associated with the normalisation of typical laboratory parameters of disease flare, including CRP, ferritin and liver function tests. In other studies, remission was defined as the absence of articular, systemic and laboratory evidence of disease activity, for at least 2 consecutive months, regardless of the administered therapy [53]. Finally, remission was once described as the absence of disease activity according to laboratory parameters and global assessment by the physician in charge of the patient [86].

Some publications included treatment criteria in the definition of remission. In a 24-week study assessing the efficacy of anakinra in AOSD, the primary endpoint was the achievement of remission, defined as patient afebrile (≤ 37°C body temperature), in the absence of therapy with nonsteroidal anti-inflammatory drugs (NSAIDs) in the previous 24 hours, a decrease of CRP and ferritin to reference limits and normal swollen and tender joint counts [76]. Another retrospective study defined remission as the absence of clinical features of AOSD, together with normalisation of inflammatory markers, and a reduction of the glucocorticoid dosage of at least 50% for at least 2 months [61].

Authors sometimes referred to RA’s definition of remission to assess remission in AOSD. In three retrospective observational studies, mainly characterised by a predominant joint involvement, a combination of the DAS28 remission (DAS28 < 2.6) and the disappearance of systemic features was used [33,34,39]. This definition of remission was also adopted in a clinical trial assessing the efficacy of canakinumab [28].

Partial remission

In some of the abovementioned studies, the authors introduced the definition of “partial” remission in patients not achieving the “complete” one [29,31,[48], [49], [50],67,82,83]. Partial remission was also defined as partial improvement of findings noted by the physician in charge of the patient but with the persistence of some systemic, articular or laboratory features [48,67].

Long-term remission

In several retrospective studies, sustained remission was defined as achievement and maintenance of remission over time [3,47,[49], [50], [51], [52],56,59,62,64,66,86,87], in line with the AOSD ‘monocyclic pattern’ proposed by Cush et al [4]. This outcome was achieved in approximately 30% of patients. Long-term remission was specifically defined as the complete disappearance of systemic symptoms and normalisation of laboratory evidence of disease activity for at least 1 year, up to the end of follow-up [66].

Relapse

Relapse was defined as the reappearance of clinical features or an escalation in drug administration [47,49,57,86,87]. In 1 study, authors described separately systemic and articular flares, combining clinical and laboratory criteria [47]. In another publication, relapse was defined by at least 3 out of 4 criteria newly appeared, including fever, arthralgia, skin rash and high CRP [57]. Active disease status according to the physician’s global assessment was sufficient for some authors [86].

The definition of relapse combining clinical features and reintroduction of drugs was also proposed [3,29,55,62,64]. In these studies, relapse was defined as the worsening of disease activity associated with the need for any additional therapy and/or an increased dosage of drugs, mainly glucocorticoids [3,39,55,[62], [63], [64]].

The outcome of life-threatening complications

In a retrospective study assessing the outcome of life-threatening complications, authors assessed patients with AOSD who were admitted to the intensive care unit. The outcome was described as leaving the intensive care unit without further treatment [60].

The candidate variables for the future criteria for the assessment of disease activity are summarised in Table 6.

Table 6.

Summary of the candidate variables for the future criteria for the assessment of disease activity, identified by the systematic literature review

Clinical variables	Biological	Criteria of response	Criteria of remission	Criteria of relapse	Patient-reported outcomes
• Fever • Cutaneous rash • Joint involvement  o Arthritis  o Arthralgia • Sore throat • Myalgia • Hepatomegaly • Splenomegaly • Lymphadenopathies • Pleuritis • Pericarditis	Routine and/or classical laboratory parameters • Leukocytes • Polymorphonuclears • Haemoglobin • Platelets • ESR • CRP • Ferritin • Glycosylated ferritin • Transaminases Nonroutine immunological laboratory parameters • IL-1β • IL-1Ra • IL-6 • IL-18 • TNF-α • S100A12 protein • S100A8/A9 protein • SAA protein	Definition of response in RCT of tadekinig alfa (Gabay 2018) Early predictive criteria at 3 weeks:  • Normalisation of temperature AND  • Decrease > 50% of the baseline  CRP levels OR normalisation of  CRP. Response to therapy at 12 weeks:  • ≥ 20% TJC + SJC* AND  • ≥70% decrease of CRP OR normalisation of ferritin	Remission considering clinical and biological features Remission defined as the disappearance of both clinical and laboratory signs of the disease.	Relapse considering clinical and biological features The reappearance of clinical and/or biological disease manifestation after achieving remission.	• SF36, physical and mental health domains • Patient global assessment (Visual Analogue Scale) • Patient assessment of pain (Visual Analogue Scale) • Patient assessment of fatigue (Visual Analogue Scale) • HAQ = Health assessment questionnaire to evaluate disability
			Remission considering clinical and biological features, and medication Remission defined as:  • Disappearance of clinical and laboratory signs of the disease,  • Combined with a reduction of concomitant therapy.	Relapse considering clinical and biological features, and medication • The reappearance of clinical and/or biological disease manifestation after achieving remission AND • Concomitant restart of glucocorticoids, and/or the need of any additional treatment and/or any increased dosage of drugs and/or readministration of previously stopped therapy.
		Definition of response in RCT of tocilizumab (Kaneko 2019) The ACR score (ACR 20, 50, 70) combined with systemic feature score. Systemic feature score = 5 clinical and 5 laboratory parameters.	Remission considering EULAR definitions of remission EULAR remission defined as DAS28 < 2.6
			Long-term remission, achievement of monocyclic pattern⁎⁎⁎⁎ The achievement and maintenance of remission during the follow-up/
		Systemic score⁎⁎ modified by Rau et al⁎⁎⁎	Partial remission Partial remission was defined as the achievement of not complete remission.
		DAS 28
		ACR criteria
		Physician’s global assessment (Visual Analogue Scale)

ACR, American College of Rheumatology; DAS 28, Disease Activity Score on 28 joints; EULAR, European League Against Rheumatism; HAQ, health assessment questionnaire; SJC, swollen joint count; TJC, tender joint count

^⁎According to a 44-joint assessment.

^⁎⁎Pouchot J et al. [7] Medicine (Baltimore). 1991;70:118-136.

^⁎⁎⁎Rau M et al. [81] J Rheumatol. 2010;37:2369-2376.

^⁎⁎⁎⁎Cush JJ, et al. Arthritis Rheum 1987;30(2):186-194. [4]

DISCUSSION

This systematic review assessed the available literature about clinical, radiological and biological features; previous definitions of disease activity; and remission to derive potential candidates to be included in specific criteria for the assessment of disease activity in AOSD.

Overall, the quality of most of the obtained data is remarkably low, and no consistent attempts have been made to validate the proposed measures or scores for AOSD in different cohorts. First, of the 63 articles retrieved, the majority were observational (n=59) and retrospective (n=58) studies, with poor overall quality (58 were rated poorly). Only 4 studies were controlled clinical trials. Second, clinical, biological and morphological features were assessed in a very heterogeneous way. The clinical characteristics that were most often reported were fever, cutaneous rash, arthralgia and/or arthritis, hepatosplenomegaly, and sore throat, whereas the most frequent laboratory parameters were leukocyte number, ESR, CRP and ferritin level. The definition of response, remission and relapse, as well as the assessment of disease activity and patient-reported outcomes, was also very heterogeneous.

This study has strengths and weaknesses. Although we have performed an SLR using three major databases, information on candidate variables was synthesised descriptively. Indeed, owing to the vast heterogeneity of the included studies and the overall poor quality, a meta-analysis was not considered appropriate, as recommended by the PRISMA guidelines [22] and Cochrane collaboration [88]. Furthermore, comparison of different study designs is difficult [89]. The time points for the assessment of each parameter were also very heterogeneous among the studies. Whenever there were several time points reported, only the last follow-up data were considered for analysis of clinical and morphological variables, whereas for laboratory parameters, we chose the best value (ie the strongest variation from baseline) during the course, independently of the time point, for analysis. At first glance, these differences in analysis strategies might be a source of bias. However, this issue may be minor in the present study because we did not perform a meta-analysis but only described outcomes. Although we used a PICO scheme to design the literature search strategy, we focused on the features that were used for outcome measurement in AOSD in this research but not on the effects of certain treatments. Furthermore, as our purpose was to identify potential candidate variables that could be included in the criteria for the assessment of disease activity, we sought to identify the greatest response to change for laboratory parameters. Indeed, a useful biomarker for AOSD should respond to the SMART criteria and must be sensitive and specific, measurable (with a high degree of precision), available and affordable, responsive and reproducible in a timely fashion [90]. Another potential limitation of our study might have been a potential overlap between different cohorts in 6 retrieved articles [29,[40], [41], [42],72,77]. However, this issue may be irrelevant here because we did not perform a meta-analysis. In this context, 2 papers dealt with the same cohort but with different outcomes [29,72]. Furthermore, 2 articles from the same group of authors seemed to deal each time with the same cohort but with an implemented number of patients in the most recent article [[40], [41], [42],77]. Finally, great heterogeneity was noted in the outcomes used, which reflects the current status of outcomes research in AOSD [2]. This heterogeneity is consequently associated with some deficiencies in reporting the retrieved findings as a common issue when dealing with rare diseases lacking specific measures of outcomes.

Despite the severity of the disease, standardized outcome measures are still lacking and an international agreement concerning disease activity assessment, definition of refractory patients as well as of remission, is still missing [2]. Therefore, individual empirical definitions of these features are usually used by each study, due to the lack of validated clinical tools. This might partly explain the fact that RCTs are scarce in AOSD, whereas observational studies and case reports are multiple. Limited experience with controlled clinical trials in this indication and the lack of evaluated outcome measures hamper the development of new compounds for treating AOSD [1,2]. One way to overcome these issues would be to extrapolate from established approaches in systemic onset juvenile idiopathic arthritis [14].

Another approach would be to use scores from monogenetic autoinflammatory diseases such as Auto-Inflammatory Disease Activity Index (AIDAI). However, whether the same trial designs and scores are applicable to the adult cohort of Still’s disease is an unresolved issue. Furthermore, the different manifestations, with predominant systemic or articular features, must be considered [91]. Finally, the establishment of an international consensus on tapering and discontinuation of treatment in AOSD is also limited by the lack of validated clinical tools [2].

Nevertheless, this review gives an overview of feasible data on disease characteristics and allows a first selection of relevant signs and markers for follow-up evaluation. Regarding the potential candidate variables that could be included in the criteria for the assessment of disease activity, the more common clinical characteristics were skin rash, arthralgia and/or arthritis, hepatosplenomegaly and sore throat. It is noteworthy that 3 of them are cardinal signs of the disease [55,92] and that they improved under treatment. Other clinical parameters were less often reported, especially serositis (pleuritis or pericarditis) which are yet potentially serious complications of the disease. Life-threatening complications were not reported in most of the retrieved studies, which could be explained by an overall good prognosis of the disease or more likely by preselection of less severe cases [7,85]. In this context, it should be discussed, whether even two different scores are required for AOSD to capture the articular as well as systemic manifestations. In fact, the obtained outcome measure needs to be evaluated for applicability in the full spectrum of disease.

‘Classical’ laboratory parameters were the most often reported, especially leukocyte number, ESR, CRP and ferritin level. Although the percentage of polymorphonuclears and the level of glycosylated ferritin are a cardinal sign for the former [55,92] and an important biomarker for the diagnosis of the latter [93,94], they were very infrequently reported. All laboratory parameters decreased with treatment and clinical response, except haemoglobin which increased. IL-1Ra increased immediately after treatment but seemed to decrease with time, which is consistent with similar findings in RA treated with anakinra [95]. The study by Yamada et al [69] showed that some biological parameters like ESR, CRP or ferritin are more likely to be modified and to remain low (or lower than expected) in case of flare under treatment with tocilizumab [69], whereas other markers such as WBC and LDH seem more “independent” and representative of disease activity under this therapy, which is consistent with other reports [11,96].

This systematic review shows that many criteria of response have been developed for AOSD, showing the lack of agreement on this issue. The proposed measures of outcome derived from RA (notably DAS28) did not fully evaluate the disease activity in AOSD because these mainly focus on articular involvement. Other clinical manifestations of AOSD (fever, skin rash) are not assessed by such scores, neither are other laboratory parameters of lesser use in RA (such as leucocyte count or ferritin levels). Moreover, it is unclear in AOSD, in opposition to RA, whether arthralgia is less significantly relevant than arthritis. Some works used the modified systemic score to assess the disease activity [29,54,61,74,78]. Although it is closer to AOSD disease activity, some variables, which are included in this score, could lack feasibility (eg hepatomegaly, pleuritis, pericarditis, pneumonitis). To be fully evaluated, these variables need an instrumental assessment, which could not be simultaneously available with the clinical evaluation. In addition, no clear definitions of the variables were reported and the cutoff of ferritin was arbitrarily chosen in identifying the active disease.

The available clinical trials on AOSD developed their own criteria of response, combining systemic and articular features, confirming the lack of a validated disease activity score and reducing the possible comparability and reproducibility of the results [20,27,28,76]. Thus, a validated score to accurately measure disease activity may be of importance in organizing specifically designed studies to comprehensively investigate these patients. Furthermore, different patient-reported outcomes, derived from other diseases, were used in a few studies suggesting the need to develop and validate a specific clinical tool in improving the management of these patients. This systematic review also highlights that different criteria of remission were developed for AOSD. The underlying principle of these definitions was the disappearance of signs and symptoms of the disease [48,67], but some points were pointed out. Some authors suggested to consider some more common manifestations, such as fever, skin rash and arthritis, whereas others considered laboratory parameters and the global assessment of the physician in charge of the patient [52,53,86]. Another important feature was the concomitant administration of therapies [61,76]. Some studies also reported the definition of ‘partial remission’, in patients not achieving a ‘complete remission’, proposing a possible intermediate category of disease activity that has to be considered [82,83]. Paralleling with criteria of remission, definitions of relapse were also retrieved [3,56,65,87]. The reintroduction of drugs, the appearance of clinical symptoms and the modification of laboratory tests were considered markers of flare in patients who achieved remission [65,87].

In conclusion, this review identified candidate variables that should be included in the criteria for the assessment of disease activity AOSD. The next steps will be the elaboration and validation of the score through the Delphi method [[97], [98], [99]]. A validated definition for disease activity is indeed mandatory for planning specific designed and powered studies to fully investigate these patients, allowing possible comparisons between studies and reducing the heterogeneity of the results. Such validated and accepted definitions would be of major importance not only for the clinical setting but also for facilitating the development of new drugs for the treatment of AOSD. Furthermore, our project may improve healthcare costs. Given that AOSD-related hospital charges are estimated at nearly €30,000 per patient per year during disease flares [100], better management, notably by reducing the unjustified use of expensive therapeutic strategies, could lower both direct and indirect costs.

Editor disclosure

The peer review process did not involve Editorial Board Members Bruno Fautrel and Roberto Giacomelli, and Associate Editor Tanja Stamm, and the editorial decision-making was led by editors who were not involved in the creation of this manuscript.

CRediT authorship contribution statement

Charlotte Girard-Guyonvarc’h: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation. Stéphane Mitrovic: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Data curation. Piero Ruscitti: Writing – review & editing, Writing – original draft, Visualization, Project administration, Methodology, Investigation, Data curation, Conceptualization. Bruno Fautrel: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology. Cem Gabay: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Project administration, Methodology, Data curation. Miguel Ángel Gonzalez-Gay: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Data curation. Eugen Feist: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Conceptualization. Roberto Giacomelli: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Conceptualization. Tanja Stamm: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Conceptualization.

Competing interests

CG has received research funding and consulting fees from AB2 Bio Ltd. EF has received honoraria and grant support from Novartis, Roche, Sobi. TS has received grant/research support from AbbVie and Roche, has been a consultant for AbbVie and Sanofi Genzyme, and has been a paid speaker for AbbVie, Roche, Sanofi, and Takeda. BF has received research grants from AbbVie, Lilly, MSD, and Pfizer and consultancy fees from AbbVie, Amgen, Biogen, BMS, Celltrion, Fresenius Kabi, Galapagos, Gilead, Janssen, Lilly, Medac, MSD, Mylan, NORDIC Pharma, Novartis, Pfizer, Roche, Sandoz, Sanofi-Genzyme, SOBI, and UCB. All other authors have nothing to declare.