Abstract
Background
This multicenter study aimed to address the heterogeneity of chronic recurrent multifocal osteomyelitis (CRMO) by identifying clinical subtypes through cluster analysis, exploring clinical features, treatment approaches, and short-term prognosis to improve management of pediatric CRMO.
Methods
Data from 42 pediatric CRMO patients (47.6% male; mean age 7.87 ± 3.45 years) diagnosed between June 2018 and June 2024 were analyzed. Using cluster analysis with 17 variables, patients were categorized into phenotypic subgroups. Statistical tests assessed differences in clinical features, treatment, and outcomes. Kaplan-Meier survival analysis and log-rank tests evaluated recurrence risk and final Physician Global Assessment(PGA) scores.
Results
Patients were classified into two groups: chronic bone pain and acute systemic inflammation. Significant differences were found in fever occurrence (P = 0.002), C-reactive protein (CRP), interleukin-6(IL-6), cytokines including tumor necrosis factor-α(TNF-α) elevation (P = 0.013, 0.003, 0.029), and Hemoglobin(HB), alkaline phosphatase (ALP) reduction (P = 0.007, < 0.001). PGA scores also differed significantly (P < 0.001). Although baseline differences existed, post-treatment recurrence risk and final PGA scores showed no significant differences (P = 0.247, P = 0.211). Treatment differed only in glucocorticoid use; non-steroidal anti-inflammatory drugs (NSAIDs), disease-modifying antirheumatic drugs(DMARDs), TNF inhibitors, and diphosphonates showed no statistical differences. Both groups reached remission approximately 12 months post-diagnosis.
Conclusion
Two distinct clinical phenotypes of pediatric CRMO were identified, each achieving favorable outcomes with tailored treatments. Recognizing these phenotypes may guide clinical strategies and improve prognosis for CRMO patients.
Introduction
Chronic recurrent multifocal osteomyelitis (CRMO) is a rare, non-infectious autoinflammatory disease that typically manifests in childhood or adolescence. It has an insidious onset and is primarily characterized by bone and joint pain along with recurrent fever, with a predilection for bone involvement. While the overall prognosis is favorable, delayed or untreated cases may lead to bone destruction and even pathological fractures [1]. The prevalence of CRMO in children is estimated to be 0.4–2 per 100,000 [2].
Current research highlights the heterogeneity of CRMO in terms of etiology, clinical manifestations, and treatment responses [3]. Clinically, variations exist in the presence, location, and nature of bone pain among patients [4], and some children experience persistent fever. CRMO may occur in isolation or in association with other conditions such as palmoplantar pustulosis, psoriasis, inflammatory bowel disease, inflammatory arthritis, or other rare inflammatory disorders [5]. Responses to non-steroidal anti-inflammatory drugs (NSAIDs) are inconsistent [6], and individual differences in efficacy and tolerance have been observed with newer treatments such as tumor necrosis factor-α inhibitors(TNFi) [7]– [8]. Prognostic outcomes also vary. Some patients achieve rapid remission while others require prolonged, complex treatment regimens and face a risk of recurrence [9]. These variations in clinical phenotypes pose challenges to the diagnosis, treatment, and prognosis of CRMO.
This study aims to systematically explore the diverse clinical characteristics of pediatric CRMO, classify these characteristics, identify distinct clinical subtypes, and determine their associations with treatment responses and prognostic outcomes. These findings are expected to contribute to precision medicine approaches for better disease management.
Cluster analysis, a robust tool for stratifying heterogeneous diseases into subtypes, has been widely applied in conditions like antiphospholipid syndrome, juvenile dermatomyositis, and systemic lupus erythematosus [10–12]. These studies have clarified relationships between clinical phenotypes and prognosis. In this study, we pioneer the application of cluster analysis to CRMO, aiming to reveal the associations between phenotypes and prognostic outcomes. Identifying distinct subgroups with varying clinical presentations could also provide valuable insights into the pathogenesis of CRMO and guide future research.
Patients and methods
Data collection and population
Forty-two children diagnosed with CRMO in the rheumatology and immunology departments of our hospital and its collaborating units between June 2018 and June 2024 were included in this study. The study was approved by the hospital’s medical ethics committee (approval number: SHERLLM2021011), and informed consent was obtained from the guardians of all participants.
Inclusion criteria: (1) 1. Age ≤ 18 years. (2) Disease duration ≥ 2 weeks. (3) Diagnosis meeting the Jansson criteria and/or Bristol criteria [13]– [4]. Exclusion criteria: Children with tumors, infections, immunodeficiency, or monogenic autoinflammatory diseases were excluded from the study [14].
Research methods
Data collection and Follow-Up
Data were collected using the electronic medical record system of our hospital, including the medical records, laboratory test results, imaging findings, treatment details, and follow-up information of the study subjects to summarize the clinical characteristics. A retrospective analysis of clinical data was conducted, including symptoms, physical examinations, laboratory test results, and imaging findings. Follow-up was performed retrospectively through inpatient and outpatient visits until December 2024.
Data collected
The data collected included gender, age, clinical manifestations, complications, recurrence status, past medical history, treatment methods and efficacy, follow-up duration, laboratory results, bone marrow cell morphology, bone biopsy histopathology, etiological findings, family pedigree whole exome sequencing (WES), and imaging results such as X-ray, Computed Tomography(CT), Positron Emission Tomography/Computed Tomography(PET/CT), Magnetic Resonance Imaging(MRI), and whole-body bone scans.
Clinical observation indicators and examination methods
(1) Laboratory Indicators: White blood cell count (WBC), Hemoglobin(HB), neutrophil proportion, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), ferritin, procalcitonin, human leukocyte antigen B27 (HLA-B27), anti-cyclic citrullinated peptide antibody (CCP), anti-keratin antibody (AKA), antiperinuclear factor (AFP), rheumatoid factor (RF), antinuclear antibody (ANA), and cytokines including tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), interleukin-8(IL-8), interleukin-10(IL-10), interleukin-1β (IL-1β), and interleukin-2 receptor (IL-2R). In addition, alkaline phosphatase (ALP) and serum calcium (Ca) levels were measured.
(2) Imaging Indicators: MRI and CT of the lesion sites, PET, and whole-body bone scans.
Follow-up and outcomes
Patients underwent follow-ups every 3 to 6 months in outpatient clinics and/or during hospitalization. New events, including recurrence or relapse and mortality, as well as laboratory test results, were documented. Updated follow-up data were obtained through telephone contact with patients.
The primary endpoint was defined as the modified Physician Global Assessment (PGA) during follow-up, scored as follows: (1) Fever: Body temperature ≥ 38 °C scored 1 point; no fever scored 0 points. (2) Bone pain, joint pain, or physical dysfunction: Presence scored 1 point; absence scored 0 points. (3) Elevated inflammatory markers: CRP > 8 mg/L and/or ESR > 20 mm/h scored 1 point; normal levels scored 0 points. (4) Composite Disease Activity Score: No activity scored 0 points; mild activity scored 1 point; moderate activity scored 2 points; severe activity scored 3 points [15, 16].
Definition of Remission and PGA Score Correlation: Clinical remission was defined as follows: complete remission required absence of subjective symptoms (bone pain, swelling, fever) and objective inflammatory signs (local redness/swelling, limited mobility), with normalized inflammatory markers (CRP, ESR). Partial remission was defined as ≥ 50% overall clinical improvement reported by patients/physicians, with significant reduction in inflammatory markers from baseline but incomplete normalization. No remission was categorized as unchanged or worsened symptoms with persistent biomarker abnormalities [7].The Physician Global Assessment (PGA) score, a scale reflecting physician-rated disease activity, correlated with remission states: PGA = 0 denoted complete remission; PGA reduction ≥ 50% from baseline (but > 0) indicated partial remission; PGA reduction < 50% or score increase defined no remission.
Recurrence or Relapse Evaluation: A recurrence or relapse was defined as the reappearance of clinical symptoms, an elevation in inflammatory markers, or new or expanding lesions observed on imaging, occurring after a relatively stable period during treatment.
Statistical methods
We summarized demographic variables, clinical symptoms, and laboratory data. Statistical analysis was performed using Statistical Package for the Social Sciences(SPSS) version 27.0. Normally distributed measurement data were expressed as mean ± standard deviation (x̄ ± s), with group comparisons conducted using independent samples t-tests. Non-normally distributed measurement data were reported as medians (M) with interquartile ranges (Q1, Q3) and compared between groups using rank sum tests. Enumeration data were presented as counts (%), and comparisons between groups were conducted using chi-square (χ²) tests. A two-sided P-value < 0.05 was considered statistically significant.
Cluster Analysis: The K-prototypes algorithm [17], capable of handling both continuous and categorical variables, was employed for unsupervised clustering. Variables with more than 10% missing data were excluded, leaving 17 baseline clinical covariates for analysis. These included demographic variables (age at onset, gender), clinical symptoms (disease course, fever, rash, gastrointestinal involvement, bone destruction, number of affected sites, PGA scores), and baseline laboratory indicators (WBC, HB, CRP, ESR, IL-6, TNF-α, ALP, Ca). These covariates were preselected to represent common clinical variables readily available in routine practice. Continuous variables included disease course, number of affected sites, PGA scores, WBC, HB, CRP, ESR, IL-6, TNF-α, ALP, and Ca, while the remaining variables were binary. Patients missing critical clinical covariates were excluded from the analysis.
The silhouette width method was used to determine the optimal number of clusters. When examining cluster numbers ranging from K = 2 to K = 10, the highest average silhouette width was observed at K = 2 clusters. Clustering was performed using the clustMixType package (version 0.4-2) in R (version 4.3.3). All P-values were derived from two-tailed tests, and a P-value < 0.05 indicated statistical significance.
We applied Kaplan-Meier (KM) survival curves to examine the relationship between CRMO phenotype groups and the PGA. The follow-up period began on the date of diagnosis for time-to-event analysis. Differences between groups were assessed using the log-rank test.
Results
Baseline characteristics
General information and clinical manifestations
As illustrated in Figs. 1 and 55 patients completed their initial visit and provided informed consent. Eleven patients were excluded due to missing data, and 2 patients were lost to follow-up. Consequently, a total of 42 patients (47.6% male; mean age 7.87 ± 3.45 years) were included in the analysis. Baseline characteristics are summarized in Table 1, which includes key demographic variables and clinical features. The median time from disease onset to diagnosis was 5.0 months (IQR 1.4–19.3 months). The median follow-up duration was 15.5 months (IQR 7–21 months), and the median number of bone lesions was 5.19 ± 3.46. No patient had a significant family history of CRMO or other rheumatic diseases.
Fig. 1.
Flow diagram of the study
Table 1.
General information and clinical manifestations
| Characteristic | Value |
|---|---|
| Male patients, n (%) | 20 (47.6) |
| Mean age at onset, years, x¯±s | 7.87 ± 3.45 |
| Mean number of bone lesions, x¯±s | 5.19 ± 3.46 |
| Median time to diagnosis, months (range) | 5.0 (1.4, 19.3) |
| Median follow-up time, months (range) | 15.5 (7.0, 21.0) |
| Clinical symptoms, n (%) | |
| Initial symptoms, n (%) | |
| Arthralgia | 36 (85.7) |
| Bone destruction | 31 (73.8) |
| Fever | 23 (54.8) |
| Soft tissue swelling | 22 (52.4) |
| Bone pain | 19 (45.2) |
| Bone swelling | 15 (35.7) |
| Joint effusion | 10 (23.8) |
| Bilateral bone involvement | 28 (66.7) |
| Comorbidities, n (%) | |
| Arthritis | 10 (23.8) |
| Uveitis | 0 |
| Rash | 8 (19.0) |
| Psoriasis | 0 |
| Palmoplantar pustulosis | 2 (4.8) |
| Gastrointestinal symptoms | 8 (19.0) |
Laboratory examinations
Peripheral blood analysis suggested a WBC of 7.0 (IQR 4.89–8.55) × 10⁹/L, with elevated WBCs in 12.5% (5/40) of cases. HB levels were 121.1 ± 14.5 g/L, with decreases observed in 26.3% (10/38) of patients. PLT counts averaged 377.6 ± 161 × 10⁹/L, with elevations in 68.4% (26/38) of cases. Elevated CRP levels were present in 61.9% (26/40), with a median value of 16.0 mg/L (IQR 3.1–42.0). ESR was elevated in 78.6% (33/40), with a mean value of 34.8 ± 21.9 mm/h. SF levels were elevated in 29.4% (10/34), with a median value of 70.0 (IQR 51.7–142) mg/L. IL-6 was elevated in 82.9% (29/35), with a median value of 8.55 (IQR 3.50–20.60) pg/ml. TNF-α levels were elevated in 34.3% (12/35), with a median value of 6.03 (IQR 2.44–12.50) pg/ml. ALP levels were normal in all cases (166 U/L; IQR 136–221).
Autoimmune markers revealed ANA positivity in 5.4% (2/37), RF positivity in 2.7% (1/37), and HLA-B27 positivity in 4.3% (1/23). Blood cultures were uniformly negative (26/26, 100%). WES was performed in 21.4% (9/42) of cases, revealing no mutations beyond known polymorphisms. Bone biopsies were conducted in 40.5% (17/42) of patients, all yielding positive findings: 52.9% (9/17) showed non-specific chronic inflammation and sclerosis/fibrosis; 29.4% (5/17) demonstrated sequestrum; 11.8% (2/17) had reduced hematopoietic components and granulation tissue proliferation; and 5.9% (1/17) revealed osteoporosis.
Imaging findings
Imaging results are summarized in Table 2. Abnormal CT findings were noted in 80% (20/25). MRI abnormalities were observed in all cases (42/42), including bone marrow edema in 97.6% (41/42), peripheral soft tissue swelling in 42.9% (18/42), joint effusion in 26.2% (11/42), bone destruction in 23.8% (10/42), synovial thickening in 19.0% (8/42), and muscle edema in 9.6% (4/42). PET-CT revealed abnormalities in all cases (24/24), such as uneven bone density in 41.7% (10/24), increased marrow cavity density in 29.2% (7/24), soft tissue swelling in 16.6% (4/24), bone destruction in 37.5% (9/24), hypermetabolism in 87.5% (21/24), and bone sclerosis in 29.2% (7/24). Bone scintigraphy was abnormal in all cases (8/8). The localization and number of bone lesions are shown in Fig. 2.
Table 2.
Imaging characteristics of chronic recurrent multifocal osteomyelitis(CRMO) patients
| Number (frequency, %) | |
|---|---|
| CT | 25 (59.5) |
| Abnormal, n (%) | 20 (80.0) |
| MRI | 42 (100) |
| Abnormal, n (%) | 42 (100) |
| Bone marrow edema | 41 (97.6) |
| Peripheral soft tissue swelling | 18 (42.9) |
| Joint effusion | 11 (26.2) |
| Bone destruction | 10 (23.8) |
| Synovial thickening | 8 (19.0) |
| Muscle edema | 4 (9.6) |
| PET-CT | 24 (57.1) |
| Abnormal, n (%) | 24 (100) |
| Uneven bone density | 10 (41.7) |
| Increased marrow density | 7 (29.2) |
| Soft tissue swelling | 4 (16.6) |
| Bone destruction | 9 (37.5) |
| Hypermetabolism | 21 (87.5) |
| Bone sclerosis | 7 (29.2) |
| Bone scintigraphy | 8 (19.0) |
| Abnormal, n (%) | 8 (100) |
Note.CT = Computed Tomography; MRI = Magnetic Resonance Imaging; PET/CT = Positron Emission Tomography/Computed Tomography
Fig. 2.
Localization and quantity of bone lesions in pediatric chronic recurrent multifocal osteomyelitis(CRMO) patients
PGA scores
PGA scores indicated mild activity in 11.9% (5/42), moderate activity in 54.8% (23/42), and severe activity in 33.3% (14/42).
Cluster analysis results
Cluster analysis classified patients into two groups. Figure 3 presents the factorial map of individual factors based on the two clusters. A multiple comparison of baseline characteristics between the two clusters is shown in Table 3.
Fig. 3.
Factorial map illustrating the individuals used to generate the dendrogram. Colors indicate cluster membership
Table 3.
Key characteristics of study population by cluster
| Variable, n (%) | ALL (n = 42) | cluster1 (n = 20) | cluster2 (n = 22) | p |
|---|---|---|---|---|
| Age at disease onset, years, x̄ ± s | 7.87 ± 3.45 | 7.70 ± 3.18 | 8.03 ± 3.74 | 0.762 |
| Disease course [Median (Q1, Q3)] | 5.00 (1.50, 18.50) | 13.00 (4.00, 29.25) | 1.75 (1.00, 7.00) | 0.001 |
| WBC [Median (Q1, Q3)] | 7.03 (5.06, 8.62) | 7.54 (6.48, 8.83) | 6.56 (4.12, 8.22) | 0.076 |
| HB (Mean ± SD) | 121.79 ± 4.20 | 127.80 ± 13.57 | 116.32 ± 12.69 | 0.007 |
| CRP [Median (Q1, Q3)]] | 14.55 (2.90, 40.42) | 5.61 (1.30, 16.92) | 21.50 (9.18, 42.00) | 0.013 |
| ESR (Mean ± SD) | 34.71 ± 22.75 | 29.80 ± 22.62 | 39.18 ± 22.44 | 0.185 |
| ALP (Mean ± SD) | 185.17 ± 67.66 | 243.20 ± 40.94 | 132.41 ± 36.02 | < 0.001 |
| Ca [Median (Q1, Q3)] | 2.42 (2.34, 2.47) | 2.45 (2.41, 2.48) | 2.36 (2.29, 2.44) | 0.009 |
| IL-6 [Median (Q1, Q3)] | 7.67 (3.93, 17.11) | 4.76 (2.63, 7.64) | 13.74 (7.20, 33.50) | 0.003 |
| TNFα [Median (Q1, Q3)] | 7.34 (2.44, 12.59) | 3.29 (2.44, 8.98) | 8.92 (5.07, 14.53) | 0.029 |
| Number of bone lesions [Median (Q1, Q3)] | 4.00 (3.00, 6.75) | 4.00 (2.75, 6.00) | 4.50 (3.25, 7.00) | 0.169 |
| Male, n (%) | 20 (47.6) | 8 (40.0) | 12 (54.5) | 0.527 |
| PGA, n (%) | ||||
| 1 | 5 (11.9) | 5 (25.0) | 0 ( 0.0) | < 0.001 |
| 2 | 23 (54.8) | 14 (70.0) | 9 (40.9) | |
| 3 | 14 (33.3) | 1 (5.0) | 13 (59.1) | |
| Average PGA (Mean ± SD) | 1.80 ± 0.523 | 2.59 ± 0.503 | < 0.001 | |
| Fever, n (%) | 22 (52.4) | 5 (25.0) | 17 (77.3) | 0.002 |
| Rash, n (%) | 2 (10.0) | 6 (27.3) | 2 (10.0) | 0.243 |
| Bone destruction, n (%) | 31 (73.8) | 14 (70.0) | 17 (77.3) | 0.854 |
| Gastrointestinal involvement, n (%) | 8 (19.0) | 4 (20.0) | 4 (18.2) | 0.881 |
| Bone pain, n (%) | 37 (88.1) | 19 (95.0) | 18 (81.8) | 0.346 |
Note.WBC = White blood cell count; HB = Hemoglobin; CRP = C-reactive protein; ESR = erythrocyte sedimentation rate; ALP = alkaline phosphatase; Ca = serum calcium; IL-6 = interleukin-6;TNFα = tumor necrosis factor-α ;PGA = Physician Global Assessment
Table 3 describes the demographic, clinical, and laboratory characteristics of 42 patients categorized into two main groups.
Group 1: This group included 20 patients (47.6%) with a relatively long disease course of 13.00 months (4.00, 29.25). Most patients in this group did not experience fever and exhibited normal levels of HB, CRP, and ALP. Moreover, 69.6% had a PGA score of 2. Therefore, this group was designated as the chronic bone pain group.
Group 2: This group included 22 patients (52.4%) with a shorter disease course of 1.75 months (1.00, 7.00), significantly shorter than Group 1 (P = 0.001). Fever was present in 77.3% of patients (P = 0.002). Elevated levels of CRP, IL-6, and TNF-α (P = 0.013, 0.003, 0.029, respectively) and decreased levels of HB and ALP (P = 0.007, < 0.001) were observed. In addition, 59.1% of patients had a PGA score of 3 (P < 0.001). This group had more pronounced systemic symptoms, such as fever and a higher rate of hematological involvement, and was thus designated as the acute systemic inflammation group.
No statistically significant differences were observed between Groups 1 and 2 regarding age of onset, total WBC count, ESR, rash, bone destruction, or gastrointestinal involvement (Table 3).
Treatment
In this study, 33 children were treated with NSAIDs either alone or in combination. Fourteen children received glucocorticoids in combination treatment. Twelve children were treated with DMARDs, including 10 with methotrexate (MTX) and 2 with thalidomide. Bisphosphonates were used in 13 children, and biological agents were administered to 13 children, including TNFi in 11 cases and interleukin-6 receptor monoclonal antibodies in 2 cases. Details of the therapeutic drugs used are presented in Table 4.
Table 4.
Drug treatment of pediatric CRMO
| All (n = 42) | Cluster 1 (n = 20) | Cluster 2 (n = 22) | p | |
|---|---|---|---|---|
| NSAIDs (%) | 33 (78.6) | 15 (75.0) | 18 ( 81.8) | 0.872 |
| Steroid | 14 (33.3) | 3 (15.0) | 11 ( 50.0) | 0.038 |
| DMARDs (%) | 12 (28.6) | 5 (25.0) | 7 ( 63.6) | 0.625 |
| TNFi (%) | 11 (26.2) | 7 (35.0) | 4 ( 18.2) | 0.216 |
| Diphosphate (%) | 13 (31.0) | 9 (45.0) | 4 ( 18.2) | 0.123 |
Note.NSAIDs = non-steroidal anti-inflammatory drugs; DMARDs = disease-modifying antirheumatic drugs; TNFi = tumor necrosis factor inhibitor
Table 4 indicates a statistically significant difference in glucocorticoid utilization rates between the two groups, while there were no significant differences in the use of NSAIDs, DMARDs, TNFi, or bisphosphonates. Children treated with glucocorticoids received an initial dose of 0.5–1 mg/kg/day, with a median duration of glucocorticoid treatment of 7.0 months (6.0, 9.0).
Follow-up
Prognosis across clusters
Forty-two children were followed up, with 2 lost to follow-up. The median follow-up duration was 15.5 months (IQR 7–21 months). Relapse occurred in 8 (20.0%) children, with 5 (11.9%) relapsing after drug withdrawal. At the last follow-up visit, according to the PGA-based endpoint disease state categorization, 1 patient (2.5%) was classified as having no remission (PGA > 4), 9 patients (22.5%) as having partial remission (PGA 1–4), and 30 patients (75.0%) as having complete remission (PGA = 0). Kaplan-Meier survival analysis (Fig. 4A) and log-rank tests showed no significant differences in final PGA scores (P = 0.36) or recurrence rates (Fig. 4B, P = 0.54) between the two clusters.
Fig. 4.
A. Cumulative remission survival curves for PGA in the two clusters. B. Cumulative recurrence rate survival curves for the two clusters
As illustrated by Kaplan-Meier survival analysis, no statistically significant differences were observed between the two clusters in recurrence rates or final PGA scores at the last follow-up. However, the utilization rate of glucocorticoids was significantly lower in cluster 1 than that in cluster 2. This suggests that for patients in the chronic bone pain group (cluster 1), effective disease remission and recurrence control can be achieved using NSAIDs and diphosphonates, without relying on glucocorticoids.
The median recurrence time across all cases was 6.86 ± 3.76 months, with cluster 1 showing a median of 9.0 ± 4.24 months and cluster 2 showing a median of 6.0 ± 3.67 months (P = 0.306). This indicates no significant difference in recurrence timing between the two clusters. Furthermore, for children in the acute systemic inflammation group (cluster 2), the recurrence rate following glucocorticoid tapering or withdrawal was comparable to that of cluster 1. This finding suggests that short-term, moderate-dose glucocorticoid therapy can rapidly alleviate symptoms without increasing the risk of recurrence.
Follow-Up of important indicators
For the enrolled patients, key indicators such as PGA, CRP, ESR, and MRI findings of the most severely affected site were monitored at 0, 3, 6, and 12 months post-diagnosis. MRI sequences included coronal, sagittal, and transverse long T1 and T2 signals. The lesion dimensions (left-right, anterior-posterior, and superior-inferior) were measured and summed to calculate the MRI-index, defined as the total of these three measurements.
Imaging results were independently evaluated by two experienced pediatric radiologists who assessed the extent of bone (marrow) lesions, joint surface integrity, and soft tissue swelling. In cases of disagreement, a consensus was reached through discussion.
Figure 5 illustrates the trends in PGA scores, CRP, ESR, and MRI-index at different time points. PGA scores, CRP, and ESR showed an upward trend at 6 months of treatment, although no statistically significant differences were observed between the two clusters. Both clusters achieved remission approximately 12 months after diagnosis.
Fig. 5.
Trends in PGA, CRP, ESR, and MRI-index during follow-up
Discussion
In our multicenter retrospective analysis, through the collection and organization of clinical data on pediatric CRMO and the application of unsupervised cluster analysis, we identified two distinct clinical subtypes: the chronic bone pain group (cluster 1) and the acute systemic inflammation group (cluster 2).
Different groups may reflect varying inflammatory states of the disease. In our study, CRP, IL-6, and TNF-α levels were significantly higher in cluster 2 compared to cluster 1 (P = 0.013, 0.003, and 0.029, respectively), suggesting a more obvious inflammatory state in Cluster 2. Macrophages, activated by pathogen-associated molecular patterns (PAMPs) or damage-associated molecular patterns (DAMPs), secrete large amounts of TNF-α and IL-6, as previously described [18–20]. Concurrently, T lymphocyte subset imbalances, including the activation of pro-inflammatory Th1 and Th17 subpopulations, further enhance cytokine production. In addition, changes in the bone tissue microenvironment, increased osteoclast activity, abnormal extracellular matrix remodeling, and activation of inflammatory signaling pathways such as Nuclear Factor Kappa-B(NF-κB) and Janus Kinase-Signal Transducer and Activator of Transcription(JAK-STAT )contribute to the elevated levels of these cytokines[24].
During the non-acute phase, these levels gradually normalize due to reduced inflammatory stimuli, restoration of immune cell activity, and decreased responsiveness to PAMPs and DAMPs. Anti-inflammatory mechanisms, including increased expression of cytokines like IL-10, inhibit the synthesis and release of pro-inflammatory cytokines, leading to the stabilization of IL-6 and TNF-α levels. A 2007 study by Jansson et al. similarly demonstrated varying manifestations of CRMO during different disease phases [21].
Therefore, accurately assessing the inflammatory state of patients is crucial for guiding the intensity of anti-inflammatory treatment, including decisions regarding the use of glucocorticoids and biological agents. For cluster 2, aggressive treatment with glucocorticoids and biological agents is recommended, while cluster 1 may benefit from NSAIDs or diphosphonates for anti-inflammatory therapy.
Different groups may also indicate variations in anemia status. Hemoglobin levels in cluster 2 patients were 116.32 ± 12.69 g/L, significantly lower than the 127.80 ± 13.57 g/L observed in cluster 1 patients (P = 0.007). Based on the Chinese diagnostic criterion for anemia (Hb < 120 g/L), a significant difference in anemia prevalence exists between the two groups.
The anemia observed during the acute phase of CRMO can be attributed to multiple factors. First, inflammation-mediated suppression of erythropoiesis plays a key role. During the acute phase, an imbalance in T lymphocyte subsets, including activation of T helper cell 1(Th1) and T helper cell 17(Th17), leads to the overproduction of inflammatory cytokines such as IL-6 and TNF-α [22]. IL-6 promotes hepatic synthesis of the acute-phase protein hepcidin, which binds to and degrades ferroportin, thereby reducing iron release from storage sites into the plasma. Since erythropoiesis is iron-dependent, iron deficiency inhibits red blood cell production, linking inflammation to anemia [27].
Second, bone marrow involvement in CRMO affects hematopoietic function. Inflammation can directly impact the bone marrow, the critical site for erythropoiesis. Inflammatory cell infiltration and alterations in the bone marrow microenvironment disrupt the proliferation and differentiation of hematopoietic stem cells into the erythroid lineage. Moreover, the cytokine storm within the bone marrow suppresses the activity of erythropoietin (EPO) and other hematopoietic growth factors, further impairing erythropoiesis [23].
Different subgroups may reflect distinct stages of bone metabolism in the disease. ALP levels in cluster 2 patients were 132.41 ± 36.02U/L, compared to 243.20 ± 40.94 U/L in cluster 1 (P < 0.001). The age-specific reference range for ALP is 143–406U/L, indicating reduced ALP levels in cluster 2. The decrease in ALP during the acute phase of CRMO is attributable to several factors:
Suppression of osteoblast function by inflammatory responses. CRMO patients exhibit excessive secretion of pro-inflammatory cytokines (e.g., IL-6, IL-1, TNF-α) and insufficient production of anti-inflammatory cytokines (e.g., IL-9, IL-10, IL-18) [29]. This imbalance likely plays a critical role in CRMO pathogenesis. These cytokines influence bone resorption and remodeling by activating osteoblasts and osteoclasts [30]. Since osteoblasts are a primary source of ALP, their impaired function leads to reduced ALP synthesis [31].
Imbalance in bone metabolism. During the acute phase of CRMO, bone tissue destruction increases, while bone formation remains relatively insufficient. ALP is essential for bone formation, and impaired formation reduces both the demand and production of ALP. Studies on similar inflammatory bone diseases have demonstrated a strong correlation between metabolic imbalances and fluctuations in ALP levels [24]. By linking organ damage to baseline features in this subgroup of patients, our findings add new insights to this understanding.
Treatment regimens for CRMO may differ across patient groups, emphasizing the importance of treatment stratification. Currently, CRMO treatment lacks standardization, although NSAIDs are widely recognized as the optimal first-line therapy. Unified guidelines for second-line treatment in pediatric CRMO remain scarce. The CRMO subgroup of the Childhood Arthritis and Rheumatology Research Alliance (CARRA) has proposed three standardized consensus treatment protocols for patients with inadequate response to NSAIDs and/or active spinal lesions [25].
In this cohort study, NSAIDs were the first-line treatment for all patients. For those with suboptimal responses, additional therapies, including glucocorticoids, immunosuppressants, and biological agents, were administered. Using unsupervised cluster analysis, our study identified two distinct phenotypes among CRMO patients. While no significant differences were observed in recurrence rates or final PGA scores between the two groups, the utilization of glucocorticoids was significantly lower in cluster 1 compared to cluster 2.
This finding suggests that non-hormonal anti-inflammatory treatments, such as NSAIDs and diphosphonates, effectively achieve disease remission and control recurrence in the chronic bone pain group. In the acute systemic inflammation group, the recurrence rate following glucocorticoid tapering was comparable to that of cluster 1, indicating that short-term, moderate-dose glucocorticoid therapy can rapidly alleviate symptoms without increasing recurrence risk.
These results underscore the importance of identifying and stratifying CRMO patients based on their phenotypic characteristics, as this approach can inform personalized management strategies and improve patient outcomes.
After diverse treatments for different subgroups, all displayed recurrent disease episodes but generally had a favorable prognosis. This disease is a chronic aseptic inflammation with recurrent remissions and relapses [21]. Previous studies have indicated multiple recurrences [26]. In this study, the PGA, CRP, ESR, and MRI-index of patients were monitored at 3, 6, and 12 months post-diagnosis. Compared to enrollment, the PGA scores in both groups showed statistically significant differences. PGA scores, CRP, and ESR showed an upward trend at 6 months of treatment, but no statistical differences were observed between the groups. Both groups entered remission 12 months after diagnosis. However, CRP and ESR are influenced by various factors. In this study, the MRI-index was used to assess disease activity, which partially reflects the disease status in children and offers a novel approach to evaluating disease progression.
The modified PGA score was adapted from pediatric rheumatology tools, aligning with CNO’s clinical features of bone inflammation and systemic symptoms, as seen in studies by Pardeo and Schnabel [16]– [7], who used similar composite scores integrating fever, pain, and inflammatory markers. Liu Haimei [16] also employed a comparable modified PGA, validating the relevance of these core variables in CNO assessment. However, we recognize the score lacks formal validation, developed for internal consistency in our retrospective study. As Capponi [15] noted, validated CNO scoring systems are scarce, with most studies using adapted tools.
This study has limitations. First, the sample size is small. Future research should increase the sample size and incorporate external validation in the cluster analysis, integrating new biomarkers and decision tree algorithms for more accurate phenotypic classification. Second, the two clusters identified in this study were based on clinical and laboratory characteristics at the time of CRMO diagnosis. Further statistical research is needed to explore the relationships between these factors and prognosis, as well as their dynamic changes over time. Additionally, due to the observational nature of our study, we cannot establish causal relationships between subtype classification and management strategies, such as the link between glucocorticoid tapering and bone destruction. Therefore, interventional studies are essential to determine whether clustering can guide management decisions and improve treatment outcomes, especially concerning glucocorticoid withdrawal.Furthermore, the modified PGA score, an unvalidated tool reliant on subjective physician judgment, may introduce inter-center variability and lacks radiological data (e.g., whole-body MRI), which are critical for comprehensive CNO/CRMO assessment compared to validated systems like PedCNO or RAI-CROMRIS. Finally, while whole-body MRI is standard for CRMO research, all pediatric patients in our study were unable to cooperate with this modality; instead, 24 of 42 patients underwent PET-CT for lesion counting, potentially introducing biases in lesion detection and assessment and affecting the comprehensiveness of lesion enumeration, which should be considered when interpreting results.
Conclusions
This study identified two distinct clinical phenotypes of pediatric CRMO: the chronic bone pain group and the acute systemic inflammation group. The acute systemic inflammation group had a higher rate of glucocorticoid use. Despite receiving different treatments, both groups achieved favorable clinical outcomes. Due to the significant heterogeneity of CRMO, identifying different clinical subtypes can better guide clinical practice. Although the pathogenesis and factors influencing recurrence remain unclear, future prospective studies should explore clustering-based treatment strategies.
Acknowledgements
Not applicable.
Author contributions
TY: Research design, data collection, manuscript writing; CD Yu: Research design, statistical analysis;.YC Y: Data collection, manuscript revision; WH C, BP H, M K, YJ X, DZ, ML, MW, FF W, J H: Data collection; GX S, FQ W: Guiding the research, manuscript revision; JM L: Guiding the research, manuscript revision, Funding support; J Z: Research design, guiding the research, data verification, data collection.
Funding
Beijing Research Ward Excellence Program (BRWEP2024W102100100).
Data availability
The datasets generated and analyzed during this study are not publicly available due to ethical restrictions and patient confidentiality protections. De-identified data may be made available upon reasonable request from the corresponding author, subject to approval by the institutional review boards of the participating centers. All materials and protocols used in this study are described in the manuscript, and no additional proprietary resources were utilized.
Declarations
Ethics approval and consent to participate
The study was approved by the hospital’s medical ethics committee (approval number: SHERLLM2021011), and informed consent was obtained from the guardians of all participants.
Consent for publication
Not applicable.
Competing interests
All authors declare no conflict of interest.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Tong Yue and Chengdong Yu are co-first authors of the article.
Contributor Information
Jianming Lai, Email: laijm99@sina.com.
Jia Zhu, Email: shinn13@163.com.
References
- 1.Hedrich CM, Hofmann SR, Pablik J, Morbach H, Girschick HJ. Autoinflammatory bone disorders with special focus on chronic recurrent multifocal osteomyelitis (CRMO). Pediatr Rheumatol Online J. 2013;11(1):47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Nuruzzaman F, Zhao Y, Ferguson PJ. Chronic nonbacterial osteomyelitis: insights into pathogenesis, assessment, and treatment. Rheum Dis Clin North Am. 2021;47(4):691–705. [DOI] [PubMed] [Google Scholar]
- 3.Gamalero L, Belot A, Zajc Avramovic M, Giani T, Filocamo G, Guleria S, Ferrara G, Minoia F, Hofer M, Larbre JP, Aureal M, Toplak N, Avcin T, Chighizola CB, Cimaz R. Chronic non-bacterial osteomyelitis: a retrospective international study on clinical manifestations and response to treatment. Clin Exp Rheumatol. 2020 Nov-Dec;38(6):1255–62. [PubMed]
- 4.Jansson AF, Müller TH, Gliera L, Ankerst DP, Wintergerst U, Belohradsky BH, Jansson V. Clinical score for nonbacterial osteitis in children and adults. Arthritis Rheum. 2009;60(4):1152–9. [DOI] [PubMed] [Google Scholar]
- 5.Ramachandran S, Zhao Y, Ferguson PJ. Update on treatment responses and outcome measure development in chronic nonbacterial osteomyelitis. Curr Opin Rheumatol. 2023;35(5):255–64. [DOI] [PubMed] [Google Scholar]
- 6.Schnabel A, Range U, Hahn G, Berner R, Hedrich CM. Treatment response and longterm outcomes in children with chronic nonbacterial osteomyelitis. J Rheumatol. 2017;44(7):1058–65. [DOI] [PubMed] [Google Scholar]
- 7.Schnabel A, Nashawi M, Anderson C, Felsenstein S, Lamoudi M, Poole-Cowley J, Lindell E, Oates B, Fowlie P, Walsh J, Ellis T, Hahn G, Goldspink A, Martin N, Mahmood K, Hospach T, Lj M, Hedrich CM. TNF-inhibitors or bisphosphonates in chronic nonbacterial osteomyelitis? - Results of an international retrospective multicenter study. Clin Immunol. 2022;238:109018. [DOI] [PubMed] [Google Scholar]
- 8.O’Leary D, Wilson AG, MacDermott EJ, Lowry C, Killeen OG. Variability in phenotype and response to treatment in chronic nonbacterial osteomyelitis; the Irish experience of a National cohort. Pediatr Rheumatol Online J. 2021;19(1):45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ferguson PJ, El-Shanti HI. Autoinflammatory bone disorders. Curr Opin Rheumatol. 2007;19(5):492–8. [DOI] [PubMed] [Google Scholar]
- 10.Guedon AF, Ricard L, Laurent C, De Moreuil C, Urbanski G, Deriaz S, Gerotziafas G, Elalamy I, Audemard A, Chasset F, Alamowitch S, Sellam J, Boffa JJ, Cohen A, Wahl C, Abisror N, Maillot F, Fain O, Mekinian A. Identifying high-risk profile in primary antiphospholipid syndrome through cluster analysis: French multicentric cohort study. RMD Open. 2023;9(1):e002881. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Xu L, You H, Wang L, Lv C, Yuan F, Li J, Wu M, Da Z, Wei H, Yan W, Zhou L, Yin S, Zhou D, Wu J, Lu Y, Su D, Liu Z, Liu L, Ma L, Xu X, Zang Y, Liu H, Ren T, Wang F, Du Y, Xue J, Zhang M, Tan W. Identification of three different phenotypes in Anti-Melanoma differentiation-Associated gene 5 Antibody-Positive dermatomyositis patients: implications for prediction of rapidly progressive interstitial lung disease. Arthritis Rheumatol. 2023;75(4):609–19. [DOI] [PubMed] [Google Scholar]
- 12.Ding Y, Zhou Y, Zhan F, Xu J, Duan X, Luo H, Zhao C, Yang M, Wu R, Wu L, Chen Z, Wei W, Huang C, Wu C, Zhang S, Jiang N, Xu D, Leng X, Wang Q, Tian X, Li M, Zeng X, Zhao J. Phenotypic subgroup in serologically active clinically quiescent systemic lupus erythematosus: A cluster analysis based on CSTAR cohort. Med. 2024;5(10):1266–e12743. [DOI] [PubMed] [Google Scholar]
- 13.Roderick MR, Shah R, Rogers V, Finn A, Ramanan AV. Chronic recurrent multifocal osteomyelitis (CRMO) - advancing the diagnosis. Pediatr Rheumatol Online J. 2016;14(1):47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wipff J, Costantino F, Lemelle I, Pajot C, Duquesne A, Lorrot M, Faye A, Bader-Meunier B, Brochard K, Despert V, Jean S, Grall-Lerosey M, Marot Y, Nouar D, Pagnier A, Quartier P, Job-Deslandre C. A large National cohort of French patients with chronic recurrent multifocal osteitis. Arthritis Rheumatol. 2015;67(4):1128–37. [DOI] [PubMed] [Google Scholar]
- 15.Capponi M, Pires Marafon D, Rivosecchi F, Zhao Y, Pardeo M, Messia V, Tanturri de Horatio L, Tomà P, De Benedetti F, Insalaco A. Assessment of disease activity using a whole-body MRI derived radiological activity index in chronic nonbacterial osteomyelitis. Pediatr Rheumatol Online J. 2021;19(1):123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Liu HM, Zhang TY, Ma L, Zhang ZY, Xu M, Zhang T, Xu H, Tang XM, Yang SR, Yu HG, Song HM, Sun L. Clinical features and treatment outcomes of chronic nonbacterial osteomyelitis in children: a multicenter study in China. Med J Peking Union Med Coll Hosp. 2023;14(2):278–84. [Google Scholar]
- 17.Preud’homme G, Duarte K, Dalleau K, Lacomblez C, Bresso E, Smaïl-Tabbone M, Couceiro M, Devignes MD, Kobayashi M, Huttin O, Ferreira JP, Zannad F, Rossignol P, Girerd N. Head-to-head comparison of clustering methods for heterogeneous data: a simulation-driven benchmark. Sci Rep. 2021;11(1):4202. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hedrich CM, Morbach H, Reiser C, Girschick HJ. New insights into adult and paediatric chronic Non-bacterial osteomyelitis CNO. Curr Rheumatol Rep. 2020;22(9):52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Singhal S, Landes C, Shukla R, McCann LJ, Hedrich CM. Classification and management strategies for paediatric chronic nonbacterial osteomyelitis and chronic recurrent multifocal osteomyelitis. Expert Rev Clin Immunol. 2023 Jul-Dec;19(9):1101–16. [DOI] [PubMed]
- 20.Rafferty BA, Thakrar P. Chronic Recurrent Multifocal Osteomyelitis. Med Clin North Am. 2024 Jan;108(1):227–239. [DOI] [PubMed]
- 21.Danek S, Kuzmina P, Adamcova M, Burocziova MK, Mikyskova M, Vanickova R, Kosanovic K, Pokorna S, Reinis J, Brdicka M, Alberich-Jorda T. M. Chronic inflammation decreases HSC fitness by activating the druggable Jak/Stat3 signaling pathway. EMBO Rep. 2023;24(1):e54729. [DOI] [PMC free article] [PubMed]
- 22.Jansson A, Renner ED, Ramser J, Mayer A, Haban M, Meindl A, Grote V, Diebold J, Jansson V, Schneider K, Belohradsky BH. Classification of non-bacterial osteitis: retrospective study of clinical, immunological and genetic aspects in 89 patients. Rheumatology (Oxford). 2007;46(1):154–60. [DOI] [PubMed] [Google Scholar]
- 23.Kajdaniuk D, Foltyn W, Morawiec-Szymonik E, Czuba Z, Szymonik E, Kos-Kudła B, Marek B. Th17 cytokines and factors modulating their activity in patients with pernicious anemia. Immunol Res. 2023;71(6):873–882. [DOI] [PMC free article] [PubMed]
- 24.Ganz T. Hepcidin and iron regulation, 10 years later. Blood. 2011;117(17):4425-33. [DOI] [PMC free article] [PubMed]
- 25.Weiss G, Goodnough LT. Anemia of chronic disease. N Engl J Med. 2005 Mar 10;352(10):1011-23 [DOI] [PubMed]
- 26.Hofmann SR, Schnabel A, Rösen-Wolff A, Morbach H, Girschick HJ, Hedrich CM. Chronic Nonbacterial Osteomyelitis: Pathophysiological Concepts and Current Treatment Strategies. J Rheumatol. 2016. [DOI] [PubMed]
- 27.Moreno-Mateo F, Perea SH, Onel KB. Chronic recurrent multifocal osteomyelitis: diagnosis and treatment. Curr Opin Pediatr. 2021;33(1):90–96. [DOI] [PubMed]
- 28.Manolagas SC. Birth and death of bone cells: basic regulatory mechanisms and implications for the pathogenesis and treatment of osteoporosis. Endocr Rev. 2000;21(2):115 – 37. [DOI] [PubMed]
- 29.Ralston SH. Pathogenesis of paget’s disease of bone. Bone. 2008;43(5):819–25. [DOI] [PubMed] [Google Scholar]
- 30.Zhao Y, Wu EY, Oliver MS, Cooper AM, Basiaga ML, Vora SS, Lee TC, Fox E, Amarilyo G, Stern SM, Dvergsten JA, Haines KA, Rouster-Stevens KA, Onel KB, Cherian J, Hausmann JS, Miettunen P, Cellucci T, Nuruzzaman F, Taneja A, Barron KS, Hollander MC, Lapidus SK, Li SC, Ozen S, Girschick H, Laxer RM, Dedeoglu F, Hedrich CM, Ferguson PJ. Chronic nonbacterial osteomyelitis/chronic recurrent multifocal osteomyelitis study group and the childhood arthritis and rheumatology research alliance scleroderma, vasculitis, autoinflammatory and rare diseases subcommittee. Consensus treatment plans for chronic nonbacterial osteomyelitis refractory to nonsteroidal antiinflammatory drugs and/or with active spinal lesions. Arthritis Care Res (Hoboken). 2018;70(8):1228–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kaiser D, Bolt I, Hofer M, Relly C, Berthet G, Bolz D, Saurenmann T. Chronic nonbacterial osteomyelitis in children: a retrospective multicenter study. Pediatr Rheumatol Online J. 2015;13:25. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated and analyzed during this study are not publicly available due to ethical restrictions and patient confidentiality protections. De-identified data may be made available upon reasonable request from the corresponding author, subject to approval by the institutional review boards of the participating centers. All materials and protocols used in this study are described in the manuscript, and no additional proprietary resources were utilized.