This systematic review was conducted following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analysis checklist (PRISMA) (23). The systematic review protocol was registered at the International Prospective Register of Systematic Reviews (PROSPERO) under number CRD42022351044.

The studies selected in this review involved patients, either children or adults, with cherubism and for which there was pharmacological management of the disease. The cherubism diagnosis was defined based on clinical, radiological and sometimes histological and genetical analyses (as previously summarized in (4)).

Overall, the inclusion criteria were based on the PICOS methodology (24): Population (P): cherubism subjects; Exposure (I): pharmacological management; Comparison (C): different therapies, placebo, no therapy or no comparison; Outcome (O): effective treatment; Study design (S): clinical trials, randomized studies, observational studies, case reports and case-series. Included studies needed to describe details on drug delivery and whether these treatments affected the disease evolution. No publication period or time restriction was applied. Language was limited to English and French. The following exclusion criteria were applied: 1) Studies in which only surgical management was suggested; 2) Studies on animal models; 3) Studies with no treatment; and 4) Literature reviews and personal opinions.

Literature searches were performed using PubMed (including Medline), EMBASE, Web of Science, and Scopus databases. Electronic database searches were conducted from their starting coverage date through February 1, 2022. More information on the search strategy is provided in Appendix 1 (which can be found online).

All references were managed and the duplicated hits were removed by using reference manager software (EndNote X7^® Basic-Thomson Reuters, New York, USA).

Article selection followed a two-phase process. In phase 1, two reviewers (A.E.C. & P.-E.C.) independently screened all published papers meeting the inclusion criteria in the electronic databases, using titles and abstracts. In phase 2, the same two authors independently evaluated the full text of each paper, applied the eligibility criteria, collected key information from the selected studies, and crosschecked the information. The final selection was based solely on full-text assessment of the studies. Discrepancy in paper selection led to discussion until mutual agreement. When disagreement occurred, a third author (A.L.P) was involved to make a final decision about whether to include or exclude a study.

The data collected consisted of study characteristics (authors, year of publication, country, design), population characteristics (sample size, age of participants, demographic features), methods, drug (type, dose, posology, route of administration) and outcome characteristics on cherubism (findings, timespan, follow-up and main conclusions).

The selected studies were evaluated using the Joanna Briggs Institute’s critical appraisal tools to assess risk of bias. The answer could be ‘yes’, ‘unclear’, ‘no’, or ‘not applicable’. Two reviewers (A.E.C. & P.-E.C.) independently classified the quality of each included study. In case of discrepancy, the two reviewers conciliated. Following these ratings, the risk of bias was categorized as high, if one or more criteria were not met; low, if all criteria were met; or unclear, if one or more criteria were not rigorously described (23). Figures of the quality assessment of all studies were generated using Review Manager software (RevMan v.5.3, The Nordic Cochrane Center, Copenhagen, Denmark).

A priori the following analyses were considered, and applied where appropriate: 1) Any effect on disease progression, e.g. as measured by mean and standard deviation; 2) Quantitative synthesis, including a meta-analysis (using RevMan 5.3); 3) Tests of heterogeneity using the Cochran Q test and I² statistics; 4) A fixed or random effect model based on the heterogeneity values detected, where a value greater than 50% may be considered as an indicator of substantial heterogeneity between studies.

The risk of bias across studies was assessed as an overall risk which could influence a meta-analysis. Methodological and statistical heterogeneity were evaluated by comparing the variability in study design and the risk of bias.

When the required data were not complete, the reviewers (A.E.C. & P.-E.C.) attempted to contact the study authors to obtain specific unpublished information. Three attempts were made in a 30-day period, by email to the first, second and last authors.

A summary of the overall strength of evidence available was presented using “Grading of Recommendations Assessment, Development and Evaluation” (GRADE) Summary of Findings (SoF) tables, using GRADE pro software (25).

All data, materials and methods which support the results can be found in the article or the appendices.

The initial database search identified 621 articles. After eliminating duplicate hits, 297 articles remained of which 272 were excluded after title and abstract review, leaving 25 articles for phase 2. During phase 2, nine more articles were excluded (reasons for exclusion can be found in Appendix 2 ), leaving 14 articles for qualitative synthesis. A flowchart of the process of identification and article inclusion and exclusion are shown in Figure 1 .

The 14 included studies were mostly case reports. The studies were conducted in Brazil (26), Cyprus (27), France (28), Finland (29), Israel (30), Italy (31), Japan (32), the Netherlands (33), Russia (34), Turkey (35), United Kingdom (36, 37), United States of America (38, 39). All studies were in English. The studies reported administration of various treatment options, including calcitonin, denosumab, oral bisphosphonates, adalimumab, tacrolimus and imatinib. The descriptive characteristics of the included studies are summarized in Table 1 .

In the 14 included studies the sample sizes ranged from one (26–28, 31–33, 35–38, 40) to three (39) patients, with a total of 18 subjects with cherubism. Seven patients (38.9%) were female. The mean age of the subjects was 9.6 years, ranging from 4 to 19 years ( Figure 2 ). When stated, the age at diagnosis ranged from 2 to 6 years, with a mean age of 4 years. The cherubism patients were from different ethnic origins (Caucasian (7 patients - 38,9%), Black (3 patients – 16,7%) or Asian (2 patients – 11,1%)), even if for 6 of them nothing was specified. Two patients were reported to be twins: one was a fraternal twin (with the brother having minor signs of cherubism) (30) and for the other the twin-type was not specified (31). Of the 18 patients, six (33.3%) were treated with denosumab, seven (38.9%) with various immune/inflammatory modulating drugs, and five (27.8%) with calcitonin. Table 2 summarizes the patient demographics.

In the 18 patients, diagnosis of cherubism was establish by clinical features; radiographic imaging, which allowed the cherubism severity to be graded; pathoanatomical analysis; and since 2001 (17) identification of the causative mutation, including genetic assessment of SH3BP2 mutations. For 10 patients, pathognomonic bilateral cherubism lesions were reported. Radiographic images were provided for all 18 patients. However, the severity grade and the classification used were given for only two patients (28, 35). For seven patients (38.9%), a familial history of cherubism was mentioned. In the 18 patients, two were sisters (30) and two were cousins (39). We decided to include cherubism patients with or without genetic data. Thus, for nine patients, no genetic analysis of SH3BP2 mutations was mentioned, and for one patient it was clearly stated that no genetic analysis was conducted because of the clear cherubism diagnosis (31). For eight patients, a genetic diagnosis was made, and the exact mutation was given for five patients (28, 29, 32); one study reported a mutation in SH3BP2 exon 9 but did not give the specific mutation (33). Very few extraoral cherubism manifestations were mentioned such as exophthalmia (30, 31) and dysphagia (39). Table 3 summarizes the cherubism diagnostic features of the patients.

Risk of bias was heterogeneous among the studies. Using JBI critical appraisal tools, five studies (28–30, 32, 39) were classified as low risk of bias, four (26, 31, 33, 38) as unclear risk of bias and five (27, 35–37, 40) as high risk of bias. No study fulfilled all the methodological criteria. However, as expected of case reports, all 14 articles presented clear descriptions of the patient characteristics and the treatment provided. The sources of high risk of bias included incomplete reporting of the patient history, lack of details about the method of diagnosis, incomplete post-intervention clinical condition of the patient, lack of reporting of adverse events, and absence of takeaway lessons. The complete item list is presented in Figure 3 and Appendix 3 .

A meta-analysis was not performed because of insufficient data for statistical pooling related to the type of study included; all were case reports, with a maximum of 3 patients.

A summary of the data collected concerning the treatments given to the 18 cherubism patients is presented in Table 4 . Interestingly, the number of drugs administered is relatively low (calcitonin, denosumab, bisphosphonates, anti-TNFα, tacrolimus, imatinib) and can be divided in three categories: anti-resorptive treatment, immunomodulating treatment, and calcitonin. If we consider those three categories individually, it is worth noting that for the same drug or class of drug, neither the dose, the cumulative dose, the mandatory co-prescription? (e.g. Vitamin D and Calcium for denosumab) or even the administration route was consistent among studies.

In most cases, the drug treatment was administrated as part of a therapeutic strategy to prepare for eventual surgery ( Table 4 ). The rationale for the choice of the drug therapy was not always clearly stated. In four articles (30, 32, 33, 37), denosumab and calcitonin were chosen because they are used in the treatment of central giant cell granuloma. Only two treatments were based on previous data (28). In one report, the treatment was off study and off label (39). Initiation of treatment was usually associated with the severity of the cherubism lesion. Plans for administration of the treatment were discussed and validated by a multidisciplinary team and/or an ethics committee for nine patients. Seven patients were subjected to light to thorough pre-treatment analyses (blood markers, etc.) ( Table 4 ). Adverse effects such as nausea, hypophosphatemia, hypocalcemia (sometimes severe (32)) were reported in eight patients. The choice of the duration and cessation of treatment was not systematically explained.

The clinical outcomes of the treatments are also a source of heterogeneity ( Table 4 ). Only for two patients (31, 37) did the authors clearly state the treatment was ineffective (anti-TNFα and calcitonin). For two more patients the treatments were said to be insufficient, as there was recurrence of the cherubism lesions (29). Regarding the age of the patients at treatment initiation, 10 of the patients were under 10 years old and four were above 13.

It was not established any risk of bias due to missing results (arising from reporting biases) in this study.

The overall quality of evidence identified using GRADE SoF tables was assessed as being very low ( Appendix 4 ), because of high inconsistency due to the different therapies, unclear outcomes, small sample size, and study designs classified as observational studies.

Cherubism is a rare bone disease, with slightly more than 500 case reports in the literature (1). The rarity of the disease explains the low number of papers that were included in the present systematic review. It is noteworthy that even though most of the included studies were case reports, there was great heterogeneity not only in the cases reported and in the treatments administered, but also in the way the cases were reported, which might reflect the diversity the journals where they were published. As all 14 articles included in this systematic review are case reports, the overall quality of evidence of the studies was categorized as very low, according to the GRADE criteria (25). The level of evidence was downgraded due to limitations in study designs, imprecision, and inconsistency because of evident heterogeneity across the studies. Another strong limitation is that the absence of clear expected outcomes at the end of the treatments on cherubism progression greatly challenged our efforts to evaluate the clinical implications that can be drawn from those studies. That deficiency and a lack of standardization for reporting renders the findings inconclusive, even the results of the five patients who received denosumab treatment. In addition, it’s not clear that all the cases reported as cherubism were in fact true cases of cherubism. Indeed, the pathognomonic features of cherubism are bilateral symmetric lesion of osteolysis without pain; however, some cases reported pain and non-symmetric lesions. This increased the heterogeneity of the case reports systematically reviewed here.

In the 14 papers we reviewed, the characteristics of the patients were clearly described ( Figure 2 ), and this revealed substantial variation among the patients in terms of sex, cherubism severity, and age at treatment initiation, thereby limiting comparisons. However, the disease natural history of every patient was not always clear, e.g., whether the cherubism was a sporadic or familial form was not systematically stated; similarly, the age at diagnosis. The diagnosis of cherubism is based on clinical features, radiographic images and histological features. If the first two were clearly given, the histological features were not always made explicit. The last piece of evidence allowing the establishment of a definitive cherubism diagnosis is identification of a mutation affecting the SH3BP2 gene (17). It is understandable that sequencing facilities might not be available in all care settings, but each report should contain a statement about whether a mutational analysis was done or not. Our analysis underscores the need for definition of a standardized set of basic information to be included in all cherubism reports going forward (see Appendix 4 ). We recommend that basic biologic assessment be done and reported before, during and after the treatment, and refer to the age and sex appropriate norms for this pediatric population.

Our analysis revealed three categories of treatments that were given to cherubism patients throughout the past 20 years or so. After Ueki’s team demonstrated that TNF-α was central in the initiation and maintenance of the cherubism phenotype in mice (41), it seemed appropriate to try to treat cherubism patients with an anti-TNFα drug such as adalimumab, as did Hero and colleagues (29) or Pagnini (31). However, the clinical outcomes from these attempts were less than convincing and the therapies were discontinued. Severe cherubism is characterized by re-localization of NFATc1 into the nuclei of those giant cells (20, 28). These observations led our team (28) to treat a patient exhibiting severe cherubism with tacrolimus in order to prevent NFATc1 translocation. This treatment appeared to stop the progression of the cherubism, but it did not resolve the disease.

Cherubism and central giant cell granulomas share some features such as fibrotic lesions containing multinucleated cells as well as their localization in bone, specifically in jaw bones (42). Accordingly, therapies used to treat central giant cell granulomas would seem to be natural therapeutic candidates for cherubism — calcitonin and denosumab might be examples. However, based on our present analysis, calcitonin may not be an effective treatment for cherubism, and this appears to be the case for many other diseases in which calcitonin has been tried. Indeed, calcitonin’s use in bone diseases is declining as new and more effective drugs are being developed (43). Nonetheless, nasal administration of calcitonin appeared as an interesting alternative route, especially for young patients (20, 28). Concerning denosumab, one can roundly question its use in cherubism patients: the drug has not been authorized for use in children; moreover, the dose used (120 mg) is the same as that given to adult patients for malignant diseases. Clearly, off-label off-study treatment of cherubism patients remains a concern. Gaining insight into the pathophysiology of cherubism is imperative to be able to offer an effective and evidence-based treatment to patients.

Related to this improvised approach, our analysis revealed that an expert multidisciplinary team or an ethics committee was involved in the treatment decision for only 9 of the 18 patients. This raises concerns about both the way treatment decisions are made and, more broadly, the way cherubism patients are cared for, as very few pediatricians, bone specialists or endocrinologists were involved or mentioned.

A critical piece of information missing from most of these studies is the severity grade of cherubism, as it was given only for two patients. The evolution of cherubism is not well understood; however, several severity classifications have been elaborated through the years, the oldest and simplest being that by Seward and Hankey (6) with three grades, and the most complex having six severity grades, some with up to five subclasses as proposed by Raposo-Amaral (7) or Motamedi (8). The most recent classification defines a new severity grade (the seventh), when cherubism is associated with other syndromes (9). The severity grade should also be considered in discussions about whether and when to initiate a treatment. Although the mechanism underlying disease resolution remains unknown, the expectation of eventual resolution raises questions concerning the age at which treatment initiation might be started. In our analysis, among 18 patients, four were above 13 years old, and two were young adults. Obviously, without a clear severity grade associated with specific age groups, it is difficult to judge the opportuneness of the described treatment. This highlights the importance of the severity grading (and the choice of classification) especially in the context of patient age. Because the latter is an important factor in the disease course, it raises the question of how natural resolution versus the effect of a drug can be distinguished.

The desire to repurpose various drugs to treat cherubism is highly understandable. However, when choosing a drug, its dose, and its way of administration, the clinician should be mindful that it is mostly children who will be treated. Decisions on ‘new’ drug treatments should be informed by the results of previously published reports. During follow-up of the effect of the treatment, imaging might be necessary. Again, the concerned population (children) should always be kept in mind, and limiting excessive radiation exposure should always guide the choice of the type of imaging. The anticipated adverse effects of drugs should be monitored, for example hypocalcemia in the case of denosumab (32, 38). The duration of the treatment was not always made explicit in the included articles, and for some, it was even suggested that the treatment was still ongoing when the paper was published. This information is essential and should be clearly stated, and follow-up until the end of the treatment should be included. Furthermore, any adverse effects, even well after treatment discontinuation, should be mentioned and bone markers should have been tested before, during and after the cessation of the treatment, but generally were not.

While the aim of the treatment was not always clearly stated, all 14 case reports sought to improve the condition of the cherubism patients. Most of the outcomes that were mentioned focused on the jaw-bone aspect, and some also concerned the lesions themselves, with reduction of the numbers of multinucleated giant cells or their activation. However, neither the outcomes nor their evaluation was standardized, and this lack of consistency should be improved.

Treatment of cherubism patients with calcitonin had various effects, from no improvement (37) to reported clear improvement (27). If we assume calcitonin has some efficacy, differences in the duration of treatment, the dose and also patient age and probably cherubism severity might explain the discrepancy.

A striking effect of treatment with denosumab is that jaw-bone sclerosis was observed irrespective of the dose, the duration, the age and probably the cherubism severity (for example, see the images in (34)). But, how satisfied should we be with transforming the soap-bubble like cavities of the jaw bone into sclerotic bone, and does this really improve the patient’s condition and the disease progression? This is highly questionable, especially in a mainly spontaneously regressive disease. In addition, apart from the known side-effects of denosumab, its long-term effects are poorly reported and the high doses given to children remain a concern. External expert oversight, even in studies granted ethical approval, needs to be improved. What is now clearly missing is clinical information about the state of the jaw bones of those patients long after the treatment. One study stated that the serum level of C-terminal telopeptide increased (30) after the discontinuation of the treatment (i.e., some bone renewal had resumed).

Alendronate alone was reported to be associated with a cessation of the progression of the lesions, but the study involved only a single patient, 13 years old (36). When combined with an anti-TNF-α, in another study, the authors concluded that the treatment was ineffective (31). Anti-TNF-α alone was also declared to be an ineffective treatment (29).

Tacrolimus treatment given to one of our own patients, although with a peculiar dosing, appeared to be effective in reducing measurable intermediary outcomes related to cherubism activity (such as a decreased number of TRAP-positive multinucleated giant cells, the nuclear location of NFATc1, and the RANKL/OPG ratio of histological samples). In this unique patient with severe cherubism, we observed a cessation of both lesion progression and osteolysis after tacrolimus administration (28). This treatment, supported by evidence of NFATc1 pathway involvement in cherubism and with measurable outcomes, paves the path to stronger studies targeting both osteoclasts and immune cells, as tacrolimus does (28).

Clinical outcomes were satisfying with imatinib, a tyrosine kinase inhibitor, as the volume of the lesions decreased (39). Interestingly, this off-study and off-label treatment efficient to decrease the volume of the patients’ lesions did not have any effect on the cherubism mouse (44). However, Yoshimoto and colleagues demonstrated that treating with a very new second-generation SYK inhibitor (Entospletinib) rescued the cherubism phenotype (45). However, mainly adult patients suffering for malignant hemopathies (such as acute myeloid or lymphoblastic leukemia), were so far included in clinical trials (46)(see clinicaltrials.gov for details). Only few cellular tests have been conducted with this molecule in cells from infant acute lymphoblastic leukemia (47, 48). So Entospletinib could pave the way to a new approach incherubism therapy, however further studies are needed in this pediatric population.

Our analysis of the 14 included articles in the present systematic review prompted us to suggest a checklist for anticipating any cherubism treatment and helping to standardize its reporting, while waiting for new knowledge on cherubism pathogenesis or results from multicentric clinical trials. The items are listed in Appendix 5 .