Classification and management strategies for paediatric chronic nonbacterial osteomyelitis and chronic recurrent multifocal osteomyelitis

ABSTRACT Introduction Chronic non-bacterial osteomyelitis (CNO) is an autoinflammatory bone disease that most commonly affects children and adolescents causing significant pain and damage to bones. The absence of diagnostic criteria and biomarkers, an incomplete understanding of the molecular pathophysiology, and lack of evidence from randomized and controlled trials make the diagnosis and care challenging. Areas covered This review provides an overview of the clinical and epidemiological features of CNO and displays diagnostic challenges and how they can be addressed following strategies used internationally and by the authors. It summarizes the molecular pathophysiology, including pathological activation of the NLRP3 inflammasome and IL-1 secretion, and how these observations can inform future treatment strategies. Finally, it provides a summary of ongoing initiatives aiming at classification criteria (ACR/EULAR) and outcome measures (OMERACT) that will enable the generation of evidence through clinical trials. Expert opinion Scientific efforts have linked molecular mechanisms to cytokine dysregulation in CNO, thereby delivering arguments for cytokine blocking strategies. Recent and ongoing collaborative international efforts are providing the basis to move toward clinical trials and target directed treatments for CNO that find approval by regulatory agencies.


Introduction
Chronic non-bacterial osteomyelitis (CNO) is a noninfectious (auto-)inflammatory bone disease that primarily affects children and adolescents [1]. The disease spectrum spans from sometimes singular mild and self-limiting bone inflammation to chronically recurring or active multifocal diseases that can cause fractures or bone deformity. Multifocal disease is also referred to as Chronic Recurrent Multifocal Osteomyelitis (CRMO) [2][3][4].
The absence of diagnostic or classification criteria for CNO/CRMO, validated and/or internationally agreed outcome measures, and limited understanding of pathological mechanisms involve complicated clinical research, especially clinical trials [2,4]. However, recent advances in understanding the exact molecular mechanisms behind CNO/CRMO and international efforts to agree case definitions and measurable outcomes promise future successful delivery of patientfocussed research [3][4][5].
This manuscript summarizes the current understanding and future directions to agree case definition, diagnostic approaches, laboratory and clinical research working toward informed and individualized treatments.

Classification criteria
Currently, there are no classification criteria for CNO/CRMO available for clinical trials and laboratory studies, but international collaborative efforts are underway. A collaborative effort between the American College of Rheumatology (ACR) and the European Alliance of Associations for Rheumatology (EULAR) aims to develop classification criteria for CNO [33]. Members of this group identified clinical and imaging features of the disease by collecting 450 cases from 7 countries, 264 of which were CNO/CRMO cases (59%), 145 (32%) were mimickers and 41 (9%) were excluded. They identified candidate items that demonstrated key features of CNO/CRMO compared to mimickers. These include female predominance, lack of fever, clavicular swelling, and symmetrical pattern of bone lesions. The candidate items underwent review by experts followed by a consensus exercise. A first draft of the resulting criteria is expected to be available by the end of 2022.
Failure to produce IL-10 and IL-19 in monocytes from CNO/ CRMO patients has been linked to altered activation of mitogen-activated protein kinase (MAPK). Reduced activation of extracellular-signal regulated kinases (ERK) 1 and 2 contributes to impaired phosphorylation of the transcription factor Signalling protein (Sp-)1, its nuclear shuttling and recruitment to the IL10 and IL19 promoter regions [34]. At the same time, reduced ERK1/2 activity results in impaired phosphorylation of histone H3 at serine 10 (H3S10p), an 'opening' epigenetic mark that allows transcription factor binding and gene expression [34,35]. Together, these molecular alterations result in a pronounced reduction of IL-10 and IL-19 expression in monocytes from CNO/CRMO patients.
Altered expression of immune-regulatory IL-10 also plays a role in the pathophysiology of IBD. Reduced IL-10 expressions due to polymorphisms in the promotor region are also a proposed pathophysiological mechanism in IBD [38] and IL-10 deficient mice developed chronic enterocolitis [39,40]. Furthermore, rare monogenic defects that result in earlyonset IBD affect the IL-10 signaling pathway [39,41]. Associations between IBD and CNO/CRMO have been noted above (Section 3. Clinical Picture), and Schilling et al. have described similarities in the manifestation of enteropathic spondyloarthropathies with five patients diagnosed with SAPHO and IBD [29]. A link between enteropathic inflammation and lymphocytic infiltration of the synovium may exist in both conditions. Underlying mechanisms, however, remain unknown and it remains unclear what role (additional) genetic factors, such as HLA-B27, may play.
In human CNO/CRMO and induced arthritis in mice, reduced expression of IL-10 (and IL- 19) has also been linked with increased assembly of the cytoplasmic multi-protein complex referred to as the NLRP3 (NLR family pyrin domain containing 3) inflammasome [37,42,43]. After activation, NLRP3 oligomerises with the adaptor molecule ASC (apoptosis-associated speck-like protein containing a CARD ((caspase recruitment domain)) and pro-caspase-1 to form the NLRP3 inflammasome. The result of inflammasome assembly is the activation of caspase-1 that cleaves cytoplasmic inactive pro-IL -1β and pro-IL-18 into their active forms (IL-1β and IL-18) that are then released from cells. Furthermore, caspase-1 activation mediates the induction of inflammatory cell death ('pyroptosis') through gasdermin-D that promotes the release of intracellular components, including assembled inflammasomes and cytokines, which contribute to amplification and 'spreading' of inflammation [44].
Notably, in addition to increased inflammasome activation, CNO/CRMO patients also exhibit increased expression of inflammasome components and IL-1β on the RNA [45] and protein levels [42]. Increased gene expression in monocytes is associated with DNA hypomethylation of the corresponding genes (NLRP3, PYCARD [PYD And CARD Domain Containing]) [42]. Similar findings have previously been reported in cryoporin-associated periodic syndromes (CAPS), a condition caused by gain-of-function mutations in the NLRP3 [45]. In CAPS, uncontrolled IL-1β release results in systemic inflammation, recurrent bouts of fever, cold-induced urticaria, musculoskeletal symptoms and, in some patients, bony overgrowth affecting patellae [45]. While the exact molecular mechanisms are not well understood, increased inflammasome expression and assembly also play a role in CNO/CRMO [46]. Untreated CAPS patients exhibited demethylation of inflammasomerelated genes during monocyte-to-macrophage differentiation when compared to healthy controls, which was reversible in response to treatment [47]. Whether DNA demethylation and increased gene expression in CNO/CRMO are caused by ongoing inflammation and reversible by treatment or represent an early event in disease pathology of CNO/CRMO remains currently unknown.
Two individual genes were suggested to contribute to the pathophysiology of CNO/CRMO and at least partly explain increased inflammasome expression and activity: FBLIM1 and P2RX7.
Whole exome sequencing detected a homozygous mutation in the FBLIM1 gene of a child with CNO/CRMO with consanguineous parents and a compound heterozygous situation in another patient [48]. The FBLIM1 gene encodes for Filamin-binding LIM protein 1 (FBLP1) or 'migfilin', which is involved in the regulation of bone remodeling. In murine models, the loss of FBLP-1 upregulates RANKL expression by promoting ERK1/2 activation [49]. Furthermore, increased ERK1/2 activation results in increased inflammasome component expression and inflammasome activity contributing to IL-1 release [48,50]. The authors concluded that pathogenic variants in FBLIM1 variants contribute to CNO/CRMO in individual patients through increased inflammasome expression and assembly. Two recent studies tested the frequencies of FBLIM1 variants across large CNO/CRMO cohorts. While Adamo et al. reported a high frequency of rare variants of FBLIM1 in an Italian cohort of 80 patients [51], O'Leary et al. did not find enrichment of FBLIM1 variants above the healthy population in a cohort of 43 Irish patients (5/43,12%) [52]. Thus, it remains unclear whether This leads to reduced phosphorylation (activation) and nuclear shuttling of the transcription factor signaling protein (SP-)1, and reduced phosphorylation of Histone H3 at position Serine (H3s10p) at the IL10 and IL19 promoters (both located in the 'IL10 cluster' on chromosome 1), an 'opening' epigenetic mark. In CNO/CRMO patients, this results in reduced expression of the immune regulatory cytokines IL-10 and IL-19. Reduced abundance of these immune regulatory cytokines contributes to increased NLRP3 inflammasome activation, a process in which NLRP3 oligomerises with the adaptor molecule ASC and the inflammatory caspase 1 (Casp1). Caspase 1 enzymatically cleaves inactive pro-IL-1β and pro-IL-18 into their mature, active forms IL-1β and IL-18 that then get released from cells. IL-1β acts as a powerful pro-inflammatory mediator through its effects on the transmembrane IL-1 receptor (IL-1 R). Activation of IL-1 R contributes to increased expression of NFkB-dependent genes, including IL-1, but also inflammasome components such as ASC (encoded by PYCARD) and osteoclastogenesis/osteoclast activation. In patients with CNO/CRMO, there is increased expression of pro-inflammatory genes such as IL-1 and PYCARD). Recently, increased activity of the transmembrane P2×7 receptor (P2×7 R) has been linked with CNO/CRMO. Activation of P2×7 R results in K+ efflux, which stimulates inflammasome assembly, and Ca2+ influx, which contributes to MAPK activation. Notably, also increased MAPK activation results in inflammasome assembly and IL-1 activation. Recently reported rare cariants in the FBLIM1 gene lead to increased inflammasome expression and activation, also through increased ERK 1/2 activation (lack of inhibition through FBLP-1). (TLR Toll-like receptor; MAPK mitogenactivated protein kinase; ERK1/2 extracellular signal-regulated kinases 1 and 2; Sp-1 Signalling protein-1; H3S10p histone H3 at serine 10; IL interleukin; NLRP3 NLR family pyrin domain containing 3; Casp1 caspase-1; IL-1 R Interleukin 1 receptor; P2×7 R P2×7 receptor; FBLP1 Filamin-binding LIM protein 1). FBLIM1 variants play a pronounced role on the population level.
The transmembrane P2×7 receptor is a key regulator of inflammasome activity (Figure 1). Activation of P2×7 with ATP results in potassium efflux from cells and NRLP3 inflammasome assembly. Increased P2×7 activation had previously been linked to increased inflammasome activity in the context of Majeed syndrome (Box 1, below), an early-onset systemic autoinflammatory disease with nonbacterial osteomyelitis as a key feature [53]. Notably, increased expression and activity of P2×7 has been reported in a SAPHO patient who exhibited increased inflammasome assembly and release of IL-1β when compared to controls after priming of monocytes with LPS or their stimulation with ATP [54,55]. The authors furthermore showed that, in monocytes from the SAPHO patient, blockade of P2×7 fully abolished IL-1β secretion and successfully treated the patient with the recombinant IL-1 receptor antagonist [56]. Recently, rare and common damaging variants of P2RX7 have been reported in a large national cohort of CNO/CRMO patients (n = 191, Germany) [57]. Functional assays in genetically modified THP-1 monocyte cell-lines linked CNOassociated variants with increased pro-inflammatory signaling but reduced cell death/pyroptosis when compared to wildtype P2×7. This links variants in P2×7 and associated effects on downstream inflammasome assembly and pyroptosis with the pathophysiology of CNO/CRMO that may result in future patient stratification and target-directed treatments (P2×7 modulation, IL-1 blocking strategies).
While the pathophysiology of CNO/CRMO is complex and underlying mechanisms may vary between individual patients, the importance of inflammasome activation and cytokine dysregulation are beyond doubt. Several studies suggested that imbalanced expression of pro-and anti-inflammatory chemokine and cytokine expression in monocyte-derived cells are key contributors to osteo-inflammation, altered bone remodeling, and bone loss. Indeed, bone remodeling depends on tightly regulated crosstalk between osteoblasts, osteoclasts, endothelia and tissue-resident immune cells. In CNO/CRMO, as in patients with osteopenia, multicellular communication within bone is disrupted [58]. A recent study in mice suggested that monocyte dysregulation and increased expression of monocyte chemoattractant protein-1 (MCP-1) can be the effect of multiple factors, including (but not limited to) IL-1β and TNF-α [59]. Indeed, elevated monocyte activation markers, including MCP-1, have been associated with CNO/CRMO and disease activity in preliminary studies [60]. Because MCP-1 can activate intracellular signaling cascades including MAPKs and NF-κB, this may auto-amplify pathology in CNO/CRMO.
The importance of inflammasome-related pathology and monocyte dysregulation is underscored by the presence of sterile osteitis in several rare monogenic autoinflammatory diseases (Box 1).

Differential diagnosis
In the absence of prospectively validated and internationally accepted diagnostic or classification criteria, CNO/CRMO remains a diagnosis of exclusion. Important differential diagnoses are summarized in Table 1. Patients are frequently initially incorrectly diagnosed with bacterial osteomyelitis and treated with antibiotics, particularly those with unifocal disease [10,68]. Given the frequent presence of nocturnal bone pain, osteoid osteoma and malignancy must be considered. Notably, vitamin C deficiency/scurvy can mimic CNO/CRMO with progressive bone pain, signal alterations on MRI (Magnetic Resonance Imaging), and bone histology showing monocyte and lymphocyte infiltration and should be considered in patients with a restricted diet (including 'picky eaters') [2] (See Box 2). The diagnostic approach usually chosen by the authors is summarized in Figure 2.

Diagnostic criteria
To date, no internationally agreed and validated diagnostic criteria are available for CNO/CRMO. However, two sets of criteria have been suggested that can aid the diagnosis of CNO/CRMO [1,69] and potentially avoid bone biopsies in a proportion of patients (Table 2). Notably, both sets of criteria were developed in single center cohorts of less than 100 CNO/CRMO patients and have not been validated in unrelated cohorts or against a range of differential Box1. Knowledge from related monogenic diseases.
Several monogenic autoinflammatory diseases characterized by pathologically increased IL-1 signaling include noninfectious osteomyelitis as a prominent feature, including Majeed syndrome, deficiency of IL-1 receptor antagonist (DIRA) and pyogenic arthritis, pyoderma gangrenosum, and acne syndrome (PAPA). In these diseases, cytokine dysregulation results in systemic inflammation.
• Majeed syndrome is a rare autosomal recessive disorder that results in early-onset CNO/CRMO and dyserythropoietic anemia. It is caused by homozygous loss-offunction mutations in the LPIN2 gene encoding for the Lipin 2 protein [61]. Experiments in murine and human macrophages suggested that Lipin 2 regulates NLP3 inflammasome activation by affecting P2×7 receptor activation [53].
• DIRA is an autosomal recessively inherited disease that results in pustulosis, joint swelling, osteitis and periostitis starting in the first weeks of life. DIRA is caused by mutations in the IL1RN gene encoding for the IL-1 receptor antagonist, a post-transcriptional regulator of IL-1 expression that competes with IL-1α & IL-1β for the binding to the type I IL-1 receptor without activating it. Therefore, mutations in the IL1RN gene result in uninhibited signaling of IL-1α and IL-1β [4].
• PAPA follows an autosomal dominant inheritance and presents with erosive arthritis in childhood, with acne and pustulosis becoming more evident in puberty, though significantly milder, related symptoms can occur in CNO/CRMO. PAPA is caused by gain-of-function mutations in the Proline-serine-threonine phosphatase -interacting protein 1 (PSTPIP1) gene [62]. The PSTPIP1 protein interacts with pyrin to act as a negative regulator of NLRP3 inflammasome [63]. Mutated PSTPIP1 from peripheral blood leucocytes from patients with PAPA and cell-lines transfected with PAPA-associated mutants results in increased pyrinbinding and IL-1β production [64]. PSTPIP1 belongs to the Pombe Cdc15 homology (PCH) family of proteins and its paralog PSTPIP2 has been implicated in the pathogenesis of CNO/CRMO in murine models [65,66]. Loss of function mutations in PSTPIP2 in mice results in a clinical phenotype of severe CNO/CRMO and, IL-1β has been identified as playing a central role in the pathophysiology of the autoinflammatory process [67].
diagnoses/mimicker diseases. A large French cohort study found that retrospectively applying the Jansson et al. criteria [1] to a cohort of patients may have avoided biopsy in 27 out of 110 cases [12]. However, this study also did not include key differential diagnoses as a control cohort.

Laboratory findings
Most patients with CNO/CRMO display normal or mildmoderately raised inflammatory markers (ESR/CRP) [1,12,13]. The Eurofever registry (n = 485) and the German National Pediatric Rheumatologic Database (NPRD) (n = 774) are two of the largest case series in CNO/CRMO [10,70]. The Eurofever Registry found ESR was above the local normal range in 59% of patients, and CRP was elevated in 49% [10]. The German NPRD study found an elevated CRP >1 mg/dl in 18.0% of the patients and a mean ESR of 18.7 mm/h [70]. Only a minority of patients have highly elevated inflammatory markers, and the presence of this raises suspicion for infection (although it is important to note that inflammation markers can also be normal in bacterial osteomyelitis) or malignancy. When developing the 'Bristol' criteria, Roderick et al. suggest consideration of a bone biopsy in children with significantly raised CRP (>30 g/L) to exclude infection [69]. Full blood count is typically normal but can show monocytosis [9]. ANA (Anti-nuclear Antibody) and HLA-B27 (Human Leucocyte Antigen B-27) are negative in most patients [10,12]. The German NPRD study found the prevalence of HLA-B27 positive patients to be 15.3%, which is above the regional prevalence of 8-10% for HLA-B27 positivity [70]. This is  . Chronic inflammation with plasma cells (white arrow) and fibroblastic proliferation (black arrow) may be present raising a possibility of chronic osteomyelitis. Scurvy was suspected because of a restricted diet (the patient refused to eat fruit or vegetables) and confirmed by not detectable blood vitamin C levels. Substitution of ascorbic acid resulted in rapid and sustained pain relief. Taken together, in patients with chronic bone pain, it is necessary to exclude relevant differential diagnoses, such as scurvy, through history and appropriate clinical and laboratory testing. Further investigations, such as histology and MRI imaging, may show inflammatory changes that could represent either CNO/ CRMO or another systemic disease such as scurvy, leading to diagnostic challenges.
While not ready for use in routine clinical settings, a panel of serum proteins has been identified separating CNO/CRMO from differential diagnoses and reflecting treatment response. However, these require prospective validation in unrelated cohorts. A single center study by Hofmann et al. compared 25 serum proteins in 56 CNO/CRMO patients with healthy individuals, patients with Crohn's and patients with Juvenile Idiopathic Arthritis (JIA) [73]. Patients with CNO/CRMO exhibited significantly higher levels of the cytokines: interleukin-6 (IL-6) and interleukin-12 (IL-12), the chemokines: C-C motif chemokine 11 (CCL11)/ eotaxin, macrophage inflammatory protein-1b (MIP-1b), RANTES (regulated on activation normal T cell expressed and secreted), monocyte chemoattractant protein-1 (MCP-1), and the soluble interleukin-2 receptor (sIL-2 R). Multiparameter discrimination allowed differentiation between patients with CNO/CRMO and healthy controls/patients with Crohn's. The discrimination analyses used could not differentiate between those with JIA and CNO/CRMO. Some of these proteins are also correlated with disease activity scores (PedCNO scores, see below) in response to treatment. Although there were no differences in the clinical presentation at diagnosis in those who did and did not respond to standard naproxen therapy, persistently high levels of IL-12, MCP-1, and sIL-2 R correlated with incomplete remission at 6-9 and 12-15 months and may be used as markers for treatment response in the future. A later study by the same group compared a panel of 18 serum biomarkers from CNO/CRMO patients (n = 71) with patients with osteoarticular infections (n = 11), JIA (n = 62), acute leukaemia/lymphoma (n = 43), reactive arthritis (n = 7) and healthy controls (n = 59) [60]. Receiving operating characteristic (ROC) analysis identified two serum biomarkers that could be used to discriminate between CNO/CRMO, alternative diagnoses, and healthy individuals: CCL11/eotaxin and IL-6. Validation of these biomarkers in large unrelated cohorts is yet to be taken but may provide a useful diagnostic differentiator [60]. As the differences identified between cohorts with CMO/CRNO and alternative diagnoses in these studies are small, routine use in diagnosis may not prove to be realistic.
Taken together, several serum proteins tested to date correlated with a CNO/CRMO diagnosis and separated CNO/ CRMO patient samples from key differential diagnoses. While differences in their levels were small between groups and may therefore not be usable in routine clinical settings, future candidates or high-sensitivity assays may perform better. Furthermore, dynamic measurements of validated biomarker candidates may provide a helpful predictor of treatment response. However, the findings require validation in independent cohorts with additional assays.

Histology
As briefly mentioned above, lesional bone biopsies can help to exclude differential diagnoses such as infections and malignancy, particularly in the case of unifocal disease [69,74]. In CNO/CRMO, histopathological findings are not disease specific and include inflammatory cell infiltration, fibrosis, and sclerosis; even osteonecrosis can be seen, as seen in Figures 3 and 4. In some cases, likely when the affected area is missed, normal bone may be seen [10,17,75]. 'Early' lesions suggest an acute (auto-)inflammatory process with predominantly neutrophilic granulocytes and monocytes/macrophages. Later during the disease course, predominantly lymphocyte and plasma cell infiltrates can be seen, but monocytes/macrophages usually remain present [76]. The 'end-stage' of CNO/CRMO is bone sclerosis, which may only be reached after several years of inflammation [4]. However, all 'stages' of inflammation may coexist, even within the same tissue section, and therefore concomitantly show acute and chronic changes [77]. These findings suggest that an initially autoinflammatory bone disorder, over time, is characterized by (secondary) activation of adaptive immune mechanisms. Typically, CNO/CRMO patients who undergo bone biopsy have negative culture and polyvalent bacterial PCR result [75]. In a small subset of patients, positive cultures with commensal skin organisms have been identified due to bacterial contamination [17,75]. Notably, paraffin embedded bone samples should not be used for eubacterial search PCRs, as this frequently results in false-positive detection of mycobacterial nucleic acids [9,78].

Imaging
Imaging of bone lesions, their distribution, and complications are among the most important tools to diagnose and monitor CNO/CRMO. Examples of pathological changes on imaging can be seen in Figure 5 and are summarized in Table 3. Wholebody (WB)-MRI has become the gold-standard for both diagnosis and monitoring of treatment response. Bone scintigraphy has been superseded by MRI but may be considered as an alternative where WB-MRI or serial MRI imaging is unavailable or contraindicated for safety reasons. MRI is preferred because it does not involve ionizing radiation, allows assessment of bone marrow and soft tissues and has greater sensitivity at detecting the overall number of lesions compared to x-rays and bone scintigraphy [10,79,80]. The metaphyses show physiologically increased uptake on bone scintigraphy making identification of pathology at these sites very difficult and MRI has been shown to be superior in detecting metaphyseal lesions [79][80][81].
Typical MRI findings are hyperintense within bone marrow or surrounding soft tissue on T2 weighted fat saturated sequences, such as STIR (Short Tau Inversion Recovery) or TIRM (Turbo inversion recovery magnitude), sometimes associated with bony expansion as the disease progresses [72]. Longer term changes, such as damage to the physis and loss of vertebral body height in keeping with compression fractures, may also be determined on MRI. Whole body imaging is also valuable in the detection of clinically silent lesions, which may inform diagnosis (CRMO vs monofocal CNO vs monofocal bone infection) and direct intervention and/or treatment and care plans (especially in the context of vertebral involvement) [82][83][84]. Indeed, the inability to access WB imaging contributes to the delay in diagnosis [85].
Plain radiographs are frequently the first-line investigation in patients presenting with pain but often, regardless of X-ray findings, further imaging is warranted based on clinical findings. X-rays in CNO may be normal but may show signs of aggressive pathology, including sclerosis, lysis, and new bone formation [17]. Estimates of the sensitivity of plain radiographs in CNO/ CRMO varies from 31-77% [10,13], and, in cases where there are characteristic clinical symptoms and characteristic X-ray changes (such as medial clavicle expansion and sclerosis), further imaging to confirm the diagnosis may not be necessary [72]. CT is rarely used investigation of CNO/CRMO but may have a role in excluding other pathologies, such as osteoid osteoma, and in directing bone biopsy. Ultrasound is not the method of choice to detect bone lesions. However, it may be used initially to exclude extraosseous abscesses, to screen for joint effusion and synovitis in the presence of joint swelling, to image periosteal irregularities, or during the diagnostic workup while excluding differential diagnoses (such as screening of the abdomen for organomegaly or intraabdominal fluid) [17].
Two scores have been developed for use in whole-body MRI imaging in CNO: 1) RINBO (Radiologic Index for Non-bacterial Oseitis) [86] and 2) CROMRIS (ChRonic nonbacterial Osteomyelitis MRI Scoring) [87]. Both scores use characteristics such as size of lesion, bone edema, soft tissue swelling, vertebral changes, and hyperostosis to determine the level of disease activity. RINBO aimed to standardize radiological evaluation of disease burden and was found to show good correlation with clinical activity. CROMIS is a consensus-tool developed by 11 pediatric radiologists that shows good inter-rater reliability and agreement. These tools provide standardized evaluation methods for the severity of CNO/CRMO lesions on imaging, but their effect on patient management may be limited.

Time to diagnosis
Due to the nonspecific clinical, laboratory, and radiographic findings in CNO/CRMO, a delay in diagnosis is common. A retrospective survey of 284 patients in the U.S.A. reported a median duration from first symptoms to diagnosis of 2 years [85]. With increased awareness, the median time to diagnosis has decreased over time and was as low as 4 months in one recent cohort [11]. A single center study in the Southwest of England found that sending information regarding CNO/CRMO to orthopedic surgeons in the region decreased time to diagnosis [69].

Treatment and care
First-line treatment for patients with CNO/CRMO usually consists of non-steroidal anti-inflammatory drugs (NSAIDS), most commonly naproxen. A large French retrospective cohort study found NSAIDs to be as effective as first-line therapy in 97% of the patients (173 of 178) [12]. A prospective open-label study in CNO/CRMO treated 37 children with naproxen and found a significant reduction in clinically detectable lesions as well as improvement in radiological lesions, laboratory parameters and patient reported outcome measures, resulting in clinical remission of 51% of the patients after 12 months [21]. However, relapse rates to treatment with NSAIDS remain high; one study found that 50% of CNO/CRMO patients treated with NSAIDs flared within the first 2 years [18].
To develop consensus treatment plans (CTPs), a survey of members of the Childhood Arthritis and Rheumatology Research Alliance (CARRA) group was undertaken to gain opinion from clinicians on the diagnosis and treatment of CNO/CRMO. This survey, completed by 121 pediatric rheumatologists, found 95% used NSAIDs first line in children with a new diagnosis of CNO/CRMO. However, the choice of second-line treatment showed little consensus: 67% used methotrexate, 65% used TNFi, and 46% used bisphosphonates [89]. Each of these medications have been shown to be effective for use in CNO/CRMO in retrospective cohort studies [10,12,13,18,20,26,[90][91][92][93][94][95] and utilize a range of mechanisms of action to reduce the inflammatory process. Classical DMARDs, such as methotrexate and sulfasalazine, reduce the release of pro-inflammatory (including IL-1β and activation of protein C) and increase anti-inflammatory mediators [96,97]. TNFi blocks the potent pro-inflammatory cytokine TNF-α, (partially) restoring the balance between pro-and antiinflammatory (IL-10 and IL-19) cytokines [5,34,35,37]. Bisphosphonate pamidronate is a potent inhibitor of bone resorption through induction of osteoclastic inactivity and apoptosis. Furthermore, pamidronate has been suggested to restore the imbalance between pro-and anti-inflammatory cytokine expression, thereby reducing bone inflammation [98].
However, there is little comparative evidence and no prospective randomized trials comparing these treatments. Thus, treatment remains empiric and is, in most cases, not based on informed patient stratification: • A retrospective review of 71 CNO/CRMO patients across two US American tertiary care centers compared treatment modalities, categorizing treatment response as 'no response', 'partial response', or 'clinical remission' [13]. Treatment response was measured by improvement in pain, improvement in serological markers of inflammation and radiological bone healing (after >3 months). A repeated measure correlation structure was used to account for multiple treatments in the same patient. The probability of response was estimated to be 66% for sulfasalazine, 91% for methotrexate, and 91% for TNF inhibitors [13]. Notably, the authors suggested that TNFi treatment was associated with the greatest rate of clinical remission. • A larger French cohort of 178 patients had similar rates of response to treatment and reported bisphosphonate use to be effective: sulfasalazine 41% (7 of 17), methotrexate 38% (3 of 8), bisphosphonates 75% (6 of 8), and TNFi 89% (8 of 9) [12]. • A German single-center study suggested a high proportion of flares in CNO/CRMO patients treated with NSAIDs and/or short courses of corticosteroids. Favourable outcomes were seen in patients treated with bisphosphonates (pamidronate) or TNFi [18]. • The Eurofever registry cohort described found treatment with bisphosphonates and methotrexate to be superior to treatment with glucocorticoids and methotrexate [10]. No other significant difference was found between the groups [10]. • A recent retrospective review of 91 CNO/CRMO patients in the UK and Germany reported that both TNFi and pamidronate were associated with clinical and radiographic remission [19]. When compared to patients receiving TNFi, patients treated with pamidronate achieved clinical remission quicker and the number of radiological bone lesions decreased more quickly, although neither of these findings reached statistical significance. This may suggest the benefits of pamidronate in CNO/CRMO patients with vertebral involvement and preexisting structural damage. An important finding was that 19/69 (28%) of patients treated with pamidronate developed flares or failed to improve and required subsequent treatment with TNFi (only 6.9% were switched from TNFi to pamidronate). Furthermore, patients treated with TNFi experienced fewer flares than those on pamidronate. Notably, failure to respond to pamidronate is associated with female gender, a higher number of bone lesions, and highly elevated serum inflammatory markers (C-reactive Protein (CRP), Erythrocyte Sedimentation Rate (ESR)), which may aid in stratification of patients toward individual treatments.
Considering the abovementioned laboratory studies linking defects in inflammasome regulation with CNO/CRMO, IL-1 blocking strategies promise potential. However, to date, reports on the use of IL-1 blocking agents in CNO/CRMO patients are limited. This is likely caused by relatively high treatment cost associated with the absence of licensing [99][100][101][102]. In a small study including nine otherwise treatmentresistant patients, reduced systemic inflammatory markers and pain, and mixed radiological response were observed after initiation of (low-dose) anakinra treatment [1]. In a small study in adult CNO/CRMO patients, improvement of patient global scores was reported within a month of treatment. A limitation of all studies investigating IL-1 blockade in CNO/CRMO is the over-representation of complicated, otherwise treatment-refractory disease [16,101]. In the absence of comparative randomized clinical trials (RCTs), to generate evidence on the comparative effectiveness of available treatments, consensus treatment plans (CTPs) have been developed for use by the Childhood Arthritis and Rheumatology Research Alliance (CARRA). Notably, CTPs were developed based on expert consensus and are not intended as treatment recommendations. Patients with CNO/CRMO who are refractory to NSAIDs and/or have active spinal lesions can be enrolled in the CNO/CRMO international register (CHOIR) and treated following one of the three CTPs [88]: A) cDMARDs (methotrexate, sulfasalazine), B) TNFi (adalimumab, etanercept, infliximab) ± methotrexate C) Bisphosphonates (pamidronate, zoledronic acid). Use of glucocorticoids and NSAIDS may be included in each of the three regimens.

Future directions
Several tools are in development for use in clinical and research settings within CNO/CRMO, including the development of classification criteria (as discussed in Section 4) validated outcome measures and ongoing clinical trials.

Outcome measures
Effective comparison between observational studies is complicated by the lack of prospectively validated and internationally agreed outcome measures in CNO/CRMO. One outcome measure proposed by Beck et al. is the PedCNO score. This is a composite of five measures: erythrocyte sedimentation rate (ESR), number of radiological lesions, severity of disease estimated by the physician, severity of disease estimated by the patient or parent, and the childhood health assessment questionnaire (CHAQ) [21]. The score has recently been applied to the German NPRD and confirmed initially reported close correlation with patient and physician global scores [70].
In 2020, an Outcome Measures in Rheumatology (OMERACT) CNO/SAPHO working group was assembled to establish a core domain set for these conditions [103]. Unlike previous criteria, the core domain set includes input from patient stakeholders. The working group is composed of 26 clinicians/researchers and 2 patient partners. The core domain set is intended for use in clinical trials and observational studies. Items were identified through a scoping review for inclusion in a single core domain set that included CNO/CRMO and SAPHO. Qualitative research and a Delphi Process to determine the core domain set are ongoing.

Clinical trials
In the absence of licensed treatment options, randomized controlled trials comparing standard of care (e.g. pamidronate, TNFi) to 'new' agents are urgently needed to generate evidence. As patent protection has expired for medications currently used for the treatment of CNO (TNFi, pamidronate), support of studies investigating their safety and efficacy in CNO/CRMO by industry currently appears unrealistic.
However, there may be a window of opportunity to investigate trial drugs with therapeutic potential in CNO/CRMO that still have patent protection. Challenges caused by the absence of agreed case definitions, classification criteria, and outcome measures are addressed by the aforementioned international initiatives and others [70,103,104].

Expert opinion
While most patients with CNO/CRMO have favorable outcomes, a majority also face delays in diagnosis and treatment initiation that can lead to significant morbidity, including bone pain, fractures, bone deformity, leg-length discrepancy, and more. Reducing the time from symptom onset to effective treatment is one of the main challenges in the field. Improved awareness and soon to be published ACR/EULAR classification criteria (though technically explicitly proposed to define homogenous cohorts for clinical trials) will support early and correct diagnosis.
There is ongoing intensive research into optimal treatment strategies, which is supported by an increasingly better understanding of the pathogenesis of this condition. Data from retrospective single-and multi-center cohorts strongly suggest the efficacy of bisphosphonates and cytokine blocking strategies in the majority of CNO patients. Recent data from the UK and Germany suggest that patients with systemic inflammatory activity and a high number of bone lesions may benefit from TNFi more than from pamidronate, while pamidronate (as it acts faster than TNFi) may be of particular benefit for patients with vertebral involvement. However, the retrospective and not randomized character of this study (and others) does not allow us to draw final conclusions or to make clinical recommendations.
The involvement of pathological activation of inflammasomes and resulting pro-inflammatory phenotypes of immune cells delivers compelling arguments for clinical trials investigating cytokine blocking agents. In this context, the ongoing and almost concluded development of ACR/ EULAR classification criteria and the OMERACT initiative to define outcome measures are of key importance for the design of clinical medication trials. To support these efforts, a recent expert consensus initiative agreed that cytokine blocking agents should be trialed in a randomized prospective study, proposed inclusion (expert diagnosis of CNO and fulfillment of ACR/EULAR classification criteria) and exclusion criteria (skin disease requiring systemic treatment; already existing vertebral involvement with structural damage) as well as comparator drugs (pamidronate) (under review).
Finally, the development of clinical (e.g. high systemic inflammatory activity, high number of bone lesions, etc.) and molecular (gene variants, inflammasome assembly, DNA methylation, etc.) tools for patient stratification, and outcome prediction are in an early phase but promise potential for future target-directed treatments and care plans tailored to the individual CNO patients. Validation of the proposed candidates should be pursued within prospective cohort studies (such as CHOIR) or future clinical trials.
Taken together, significant progress has been made in understanding the clinical presentation of CNO, its comorbidities and complications, molecular underpinnings, and therapeutic options. However, to date, no widely accepted diagnostic or classification criteria exist, and the treatment remains empiric and has 'off-label' character. International collaborative efforts are needed to overcome the remaining hurdles.

Declaration of interest
The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.