Clinical Score for Nonbacterial Osteitis in Children and Adults

Transcription

1 ARTHRITIS & RHEUMATISM Vol. 60, No. 4, April 2009, pp DOI /art , American College of Rheumatology Clinical Score for Nonbacterial Osteitis in Children and Adults Annette F. Jansson, Thomas H. Müller, Leonhard Gliera, Donna Pauler Ankerst, U. Wintergerst, Bernd H. Belohradsky, and Volkmar Jansson Objective. To accurately differentiate nonbacterial osteitis (NBO) from other bone lesions by applying a clinical score through the use of validated diagnostic criteria. Methods. A retrospective study was conducted to assess data on patients from a pediatric clinic and an orthopedic tertiary care clinic, using administrative International Classification of Diseases codes as well as laboratory and department records from 1996 to Two hundred twenty-four patients older than age 3 years who had either NBO (n 102), proven bacterial osteomyelitis (n 22), malignant bone tumors (n 48), or benign bone tumors (n 52) were identified by chart review. Univariate logistic regression was used to determine associations of single risk factors with a diagnosis of NBO, and multivariable logistic regression was used to assess simultaneous risk factor associations with NBO. Results. NBO was best predicted by a normal blood cell count (odds ratio [OR] 81.5), symmetric bone lesions (OR 30.0), lesions with marginal sclerosis (OR 26.8), normal body temperature (OR 20.3) a vertebral, clavicular, or sternal location of lesions (OR 13.9), presence of >1 radiologically proven lesion (OR 10.9), and C-reactive protein level >1 mg/dl (OR 6.9). The clinical score for a diagnosis of NBO based on these predictors ranged from 0 to 63. A score for NBO of >39 had a positive predictive value of 97% and a sensitivity of 68%. Conclusion. The proposed scoring system helps to Annette F. Jansson, MD, Thomas H. Müller, MD, Leonhard Gliera, MD, Donna Pauler Ankerst, PhD, U. Wintergerst, MD, Bernd H. Belohradsky, MD, Volkmar Jansson, MD: Ludwig Maximilian University, Munich, Germany. Drs. Belohradsky and V. Jansson contributed equally to this work. Address correspondence and reprint requests to Annette F. Jansson, MD, Dr. von Haunersches Kinderspital, Ludwig Maximilian University, Lindwurmstrasse 4a, D Munich, Germany. Annette.Jansson@med.uni-muenchen.de. Submitted for publication August 20, 2008; accepted in revised form December 30, facilitate the diagnostic process in patients with suspected NBO. Use of this system might spare unnecessary invasive diagnostic and therapeutic procedures. In addition to general practitioners, many surgeons, including orthopedic and pediatric surgeons, as well as rheumatologists, pediatricians, infectiologists, radiologists, and other medical practitioners often encounter patients with bone pain and bone lesions in whom a diagnosis of bacterial osteomyelitis is suspected. Generally, osteomyelitis is assumed to be of bacterial origin, even in the absence of infectious agents. Infants, patients with oncologic diseases, those with immunodeficiencies, or those having experienced an injury are predisposed to the development of bacterial osteomyelitis, but increasingly, due to a move toward more differentiated radiologic techniques such as magnetic resonance imaging (MRI), increasing numbers of bone lesions are detected even among healthy children and adults. This has led to a rising prevalence of nonbacterial bone lesions, resembling bacterial osteomyelitis, that seem to have an autoinflammatory origin (1). Bone pain in patients with radiologically proven lesions, particularly those patients without associated symptoms (e.g. fever, weight loss), is a diagnostic challenge. Differential diagnoses include malignancy, benign bone tumors, bacterial osteomyelitis, and nonbacterial osteitis (NBO). Nonbacterial bone lesions have been identified in patients at nearly all ages and at any site of the skeleton. The presence of nonbacterial bone lesions is usually recognized as a criterion for inflammatory disorders, including synovitis, acne, pustulosis, hyperostosis, osteitis (SAPHO) syndrome (1 4), chronic recurrent multifocal osteitis (CRMO) (5 14), pustulotic arthro-osteitis (15), chronic sclerosing osteomyelitis, lymphoplasmacellular osteomyelitis, and other disorders. Recently, NBO was defined as a clinical entity similar to arthritis that is possibly associated with other autoinflammatory manifestations (13). Patients with 1152

2 CLINICAL SCORE FOR NONBACTERIAL OSTEITIS 1153 NBO present with uni- or multifocal bone lesions, which initially may be osteolytic but later develop into sclerotic or hyperostotic lesions, mainly affecting the metaphyses of the long bones, spine, or shoulder girdle. Twenty percent of patients with NBO have associated autoinflammatory disorders, particularly of the skin and bowel (5,7 14). The course of disease might be acute (duration 6 months), chronic persistent, or chronic relapsing (13). Complications include vertebral fractures and severe hyperostotic bone lesions. Associations of NBO with autoinflammatory diseases in 40% of first- and second-degree family members or associations of NBO among first-degree family members, as well as findings of linkage to 18q21 and data from mouse models and patients with associated Sweet syndrome support a genetic basis for NBO (8,10 13). Histopathologic findings have shown the presence of nonspecific inflammation in patients with NBO, while bacterial cultures as well as blood cultures have yielded negative results. Laboratory evidence of inflammation is present in many, but not all, patients with NBO, and patients show normal blood cell counts during episodes. Radiologically, nonbacterial bone lesions resemble bacterial bone lesions, and the occurrence of malignancy cannot be excluded. Currently, the diagnosis of NBO is based on exclusion, and validated diagnostic criteria are urgently needed (8,9,13). Evidence-based systematic investigations of nonbacterial bone lesions in large cohorts are lacking, including those in patients with SAPHO syndrome and in patients with CRMO (8,9,11,13,15,16). Due to having a suspected diagnosis of bacterial osteomyelitis, many patients with NBO are primarily treated with intravenous and/or oral antibiotics (5,7 14). Therapy for NBO, however, differs substantially from therapies for bacterial osteomyelitis and malignant or benign bone tumors. First-line treatment is nonsteroidal antiinflammatory drugs (15,16), which are effective in 85% of all patients with NBO (13). Surgical interventions are not helpful and may worsen the course of the disease (5,7). Therefore, accurate diagnostic criteria are needed for maximizing early detection of clinically significant NBO, so that patients may be spared from long-term antibiotic regimens, repeated bone resections, or chemotherapy (5,9,11,13). PATIENTS AND METHODS Design. Retrospective analysis was performed by chart review of patients older than age 3 years in whom NBO was diagnosed between 1996 and 2006 at the Dr. von Haunersches Kinderspital at Ludwig Maximilian University or the Department of Orthopedic Surgery at Grosshadern Hospital in Munich, Germany. Patients were included only if NBO could be unequivocally diagnosed according to the clinical criteria proposed by Jansson et al (13). Data retrieved included the patient s sex and age, clinical data from first presentation, including number of lesions on radiographs and number of osteolytic/sclerotic lesions, whether the reported symptoms had persisted longer than the previous 6 months, location of lesions, report of any associated disease, whether the blood cell count was normal, the levels of C-reactive protein (CRP), and family history of either palmoplantar pustolosis, psoriasis, severe acne, inflammatory bowel disease (IBD), rheumatism of the joints, NBO, or other autoinflammatory disorders. Three comparison groups, comprising patients with malignant bone lesions (group 1), those with benign bone lesions (group 2), and those with proven bacterial osteomyelitis (group 3), were used in the study, with all groups required to include age-matched patients with bone pain and radiologically proven osteomyelitic or osteomyelitic-like lesions at first presentation. Patients in groups 1 and 2 were identified from a retrospective review of 181 charts ( ) or 240 charts ( ) from the Department of Orthopedic Surgery of Grosshadern Hospital. Comparison group 3 comprised generally healthy patients who were age 3 years and had proven (on blood culture or bone specimen) nontraumatic bacterial osteomyelitis. To obtain a cohort of individuals fulfilling the postulated criteria, patients were identified by retrospective review of charts from all departments of the Klinikum Grosshadern Hospital (1,418 beds) and the Dr. von Haunersches Kinderspital (258 beds), with the exception of the neonatologic, oncologic, radiologic, and palliative medicine departments. Patients were recruited from these sites using clinical data recorded from 1993 to 2005, which were provided by Microbiological Laboratories, and additionally using diagnoses based on the International Classification of Diseases, Ninth and Tenth Revisions, from 1995 to 2006, which were provided by the Medical Controlling Centre of Ludwig Maximilian University. Statistical analysis. Univariate logistic regression was used to determine associations of single risk factors with the diagnosis of NBO, and multivariable logistic regression was used for determining simultaneous risk factor associations. An optimal multivariable model was obtained by evaluating all possible logistic regression models based on a set of candidate factors, and by selecting the model with the smallest Akaike s information criterion (an indicator of the highest predictive ability) in which all predictors included were statistically significant (P 0.05). Clinical scores for a diagnosis of NBO were developed by determining the sum of the product of the scaled logistic regression coefficients (i.e., the score coefficients) for all of the risk factors included in the optimal multivariable model. Performance operating characteristics of the clinical score were assessed using 2 measures, sensitivity (proportion of cases with the diagnosis of NBO correctly detected) and specificity (proportion of non-nbo cases [correctly] not diagnosed). All statistical computations were performed as 2-sided comparisons. P values less than 0.05 were considered significant. These analyses were performed using the open-source statistics software package R (see and the extension package ROCR (17).

3 1154 JANSSON ET AL Table 1. Distribution of risk factor associations with nonbacterial osteitis (NBO), compared with differential diagnoses, from univariate and full multivariate models* No. of patients NBO, no. (%) Differential diagnosis, no. Odds ratio for NBO (P) Bact. Mal. Ben. Univariate Multivariate All (46) Sex Male (33) (ref) 1.0 (ref) Female (56) ( 0.001) 2.5 (0.17) Age 3 11 years (62) (ref) 1.0 (ref) 12 years (30) ( 0.001) 0.45 (0.24) Number of radiologically proven bone lesions (25) (ref) 1.0 (ref) (84) ( 0.001) 12.4 (0.004) Number of lesions on bone scan (75) (ref) (41) (0.10) (76) (0.22) (81) (0.03) Missing 79 Number of lesions with marginal sclerosis (18) (ref) 1.0 (ref) (80) ( 0.001) 37.7 ( 0.001) Missing 1 Symptoms longer than 6 months No (28) (ref) 1.0 (ref) Yes (53) ( 0.001) 0.21 (0.058) Missing 1 Number of vertebral, clavicular, or sternal lesions (32) (ref) 1.0 (ref) (82) ( 0.001) 24.6 ( 0.003) Fractured vertebrae No (40) (ref) 1.0 (ref) Yes (83) ( 0.001) 1.7 (0.75) Hyperostosis No (43) (ref) 1.0 (ref) Yes (87) (0.005) 4.2 (0.29) Symmetric lesions No (41) (ref) 1.0 (ref) Yes (90) (0.001) 35.2 (0.027) Missing 1 NBO-associated diseases No (37) (ref) 1.0 (ref) Yes (85) ( 0.001) 2.6 (0.37) Blood cell count Normal (53) (ref) 1.0 (ref) Abnormal 37 1 (3) ( 0.001) 0.01 (0.008) Missing 2 CRP 1 mg/dl (36) (ref) 1.0 (ref) 1 mg/dl (59) (0.002) 13.0 (0.03) Missing 10 Body temperature Normal (51) (ref) 1.0 (ref) Elevated 32 4 (13) ( 0.001) 0.03 (0.036) Missing 1 Family history of related diseases No (41) (ref) 1.0 (ref) Yes (52) (0.12) 3.0 (0.11) Missing 4 * Bact. bacterial osteomyelitis; Mal. malignant bone lesions; Ben. benign bone lesions; ref referent; CRP C-reactive protein. Compared with differential diagnosis. Palmoplantar pustulosis, psoriasis, inflammatory bowel disease, or arthritis. Either palmoplantar pustulosis, psoriasis, severe acne, inflammatory bowel disease, rheumatism of the joints, NBO, or autoimmune diseases.

4 CLINICAL SCORE FOR NONBACTERIAL OSTEITIS 1155 RESULTS Retrospective evaluation identified 224 patients eligible for analysis; the clinical and laboratory characteristics of the patients in all groups are listed in Table 1. Of the 224 patients, 102 (45.5%) were diagnosed as having NBO, while 48 (21.4%) had malignant bone tumors (21 with osteosarcoma, 14 with Ewing sarcoma, 3 with Langerhans cell histiocytosis, and 10 with other, not necessarily primary, bone tumors), 52 (23.2%) had benign bone tumors (14 with osteochondroma, 10 with fibroma, 6 with fibrous cortical defect, 6 with juvenile bone cyst, 5 with osteoidosteoma, and 11 with other types), and 22 (9.8%) had bacterial osteomyelitis verified on bone specimen or blood culture (15 with Staphylococcus aureus, 3 with coagulase-negative Staphylococcus, and 4 with other types). Univariate associations indicated that most factors had a clinically meaningful (odds ratio [OR] 2.3), statistically significant (P 0.005) bearing on the NBO diagnosis (Table 1). Patients with 2 or more radiologically proven bone lesions had a likelihood of having NBO that was 15 times higher than that in patients with no or only 1 radiologically proven bone lesion (95% confidence interval [95% CI] 8 31, P 0.001), and patients with at least 1 lesion showing marginal sclerosis had odds of NBO 18 times higher than those in patients with no such lesions (95% CI 9 36, P 0.001). Localization of lesions in the vertebrae, clavicles, or sternum carried an increase of 10-fold in the odds of NBO (95% CI 21 49, P 0.001). Patients with symmetric lesions had odds of NBO 13 times higher than those in patients with nonsymmetric lesions (95% CI , P 0.001), and patients with the NBO-associated diseases palmoplantar pustulosis, psoriasis, IBD, or arthritis had odds of NBO that were 9.3 times higher than those in patients with no associated diseases (95% CI , P 0.001). An abnormal blood cell count and elevated body temperature are not typical symptoms of NBO but are found in other bone diseases. Accordingly, these 2 factors were associated with ORs that were below unity, i.e., an OR of 0.02 (95% CI , P 0.001) and an OR of 0.14 (95% CI , P 0.001), respectively. Increases in the CRP level are found inconsistently in all of the bone diseases considered herein, but an increased CRP level is somewhat more frequent in NBO. In this analysis, the odds of NBO increased 2.5-fold for patients with a CRP level 1 mg/dl (95% CI , P 0.002). Fractured vertebrae or hyperostosis, which are typical complications of NBO, increased the odds of NBO by 7.6-fold (95% CI , P 0.001) and 8.8-fold (95% CI , P 0.005), respectively. A retrospectively reported duration of pain in excess of 6 months increased the odds of NBO 3.1-fold (95% CI , P 0.001). The number of lesions detected on bone scan had only a weak association with NBO. No significant association with NBO was found for family history of related diseases (palmoplantar pustulosis, psoriasis, severe acne, IBD, rheumatism of the joints, NBO, or autoinflammatory disorders). Independent diagnostic values for each of the risk factors in comparison with all other risk factors were determined with multivariate ORs (Table 1). Although the associations of NBO with sex, age, symptom duration, fractured vertebrae, hyperostosis, associated diseases, and family history were statistically significant when evaluated in isolation, none of these factors had statistically significant independent diagnostic value when compared with the other risk factors (P 0.05 in all comparisons). Including all factors in a multivariable model to construct a clinical score for diagnosis is not efficient for external validation, because it causes a reduction in statistical power (due to a high ratio of predictors to sample size). Therefore, the optimal multivariable regression model, comprising 7 risk factors (listed in Table 2), was developed following the procedure described in Patients and Methods. The proposed clinical score for a diagnosis of NBO was calculated as the sum of the score coefficients. To construct an individual s score, the scaled coefficient is included in the sum when the individual has the characteristic at a diagnostic value corresponding to the scaled coefficient; otherwise, if the individual does not have the characteristic, the contribution to the score for that characteristic is zero. Therefore, the possible clinical score for a diagnosis of NBO could range from 0 to 63, with higher scores indicating a greater likelihood of NBO (Table 2). For constructing a clinical algorithm for diagnosis of NBO, operating characteristics for the clinical score were examined in terms of sensitivity and specificity. Two cutoff values stood out for defining a low-risk versus intermediate-risk versus high-risk group for NBO. A diagnostic test for NBO that defines a positive diagnosis as a clinical score exceeding 29 would capture most NBO cases (sensitivity of 97%) while falsely diagnosing 20% of non-nbo patients as having NBO (specificity of 80%). Defining the test for positivity by the higher cutoff value of 39 would lower the number of NBO cases correctly detected (68%) but decrease the false-positive rate for non-nbo cases to only 2%. Therefore, a clinical score with cutoffs at 29 and 39 was proposed as a

5 1156 JANSSON ET AL Table 2. Optimal multivariable logistic regression model for calculating the clinical score for a diagnosis of nonbacterial osteitis* Risk factor Logistic regression coefficient SE OR P Score coefficient Normal blood cell count Symmetric lesions Lesions with marginal sclerosis Normal body temperature Vertebral, clavicular, or sternal lesions Radiologically proven lesions CRP 1 mg/dl Total clinical score 63 * Risk factors were included if the regression coefficient was statistically significant (P 0.05). Score coefficients are the scaled coefficients of the logistic regression values and represent the score component for each risk factor. If the individual does not have the specific risk factor, his or her contribution to the total clinical score for that characteristic is zero. The total clinical score is the sum of score coefficients not equal to zero, with a range of OR odds ratio; CRP C-reactive protein. diagnostic tool for NBO according to the following interpretation: score values from 0 to 28 probably not NBO (nominal negative predictive value of 97%), score values from 29 to 38 uncertain diagnosis ( 80% probability of NBO), and score values of 39 probably NBO (nominal positive predictive value of 97%). DISCUSSION Little is known about autoinflammatory bone diseases. Rheumatologists and orthopedic surgeons increasingly recognize bone lesions as nonbacterial, even if well-known associated symptoms such as palmoplantar pustolosis or hyperostosis are lacking (18,19). Particularly in immunocompetent and otherwise-healthy children and adults, NBO might be much more frequent than bacterial osteomyelitis. Extensive diagnosis-related searches for patients with proven bacterial osteomyelitis from 1993 to 2006 at the Dr. von Haunersches Kinderspital and the Klinikum Grosshadern of Ludwig Maximilian University in Munich, Germany (a total of 1,676 beds) did not enable us to find 50 otherwise-healthy individuals in matching age groups fulfilling the proposed criteria. One explanation for the lack of patients in this diagnostic group might be that, in some cases, antibiotic therapies had been initiated before the patient was transferred to the hospital, so that the causative organism could no longer be detected. Thus, we had to include all otherwise-healthy individuals with proven bacterial osteomyelitis regardless of their age. Excluded from the study were children younger than age 3 years, because in infants with osteomyelitic bone lesions, bacterial osteomyelitis should be suspected primarily on the basis of anatomic findings (20). The use of universal 16S ribosomal DNA polymerase chain reaction for exclusion of patients with osteomyelitis related to Kingella kingae (21,22) and consideration of tuberculosis (23) might be helpful. Bacterial osteomyelitis and malignancy, being the 2 most important differential diagnoses with respect to NBO, require immediate attention. We herein proposed a clinical score that could facilitate the diagnosis and treatment process, especially with respect to the decision on whether to carry out invasive procedures required for diagnosis of these diseases. The clinical score, derived using quantitative statistical techniques, includes many of the factors previously associated with NBO diagnosis, such as the number, location, and symmetry of radiologically proven lesions, the presence of marginal sclerosis, body temperature, blood cell count, and CRP level (13). These results suggest that the score may be useful in diagnosing NBO. However, the performance characteristics of the clinical score were observed to be overoptimistic, since they were derived from the same data set used to optimize the score coefficients. Although this problem could be alleviated by using an optimizing model selection procedure, further external validation of the proposed clinical score is strongly needed before widespread clinical use. The clinical presentation of a patient is an important parameter that should be included in a prospective clinical score. Since only chart review was available, clinical presentation could not be reliably analyzed in this study. Patients with a suspected diagnosis of NBO usually present in a good state of health without fever, weight loss, or other signs of systemic disease (5 9,11 14,16). Patients with bacterial osteomyelitis or malignant bone lesions did show fever more often than patients with NBO. Moreover, most patients with malignancy

6 CLINICAL SCORE FOR NONBACTERIAL OSTEITIS 1157 Figure 1. Proposed diagnostic approach for nonbacterial osteitis (NBO) in an otherwise-healthy patient with bone pain. Conv. conventional; MRI magnetic resonance imaging; pts points; PPP palmoplantar pustulosis. consulted a doctor within 6 months of the onset of symptoms (data not shown). Patients with NBO showed a normal blood cell count, whereas an abnormal blood cell count was found most frequently in patients with bacterial osteomyelitis, followed by patients with malignancy. CRP values that exceeded 1 mg/dl were most frequently seen in patients with bacterial osteomyelitis and in half of the patients with malignancy. The CRP level is not a very specific diagnostic parameter. More specific parameters, such as the antinuclear antibody level, which might be elevated in NBO (13), were not available for many of the patients in this retrospective study. Some of the factors originally considered, although closely related to NBO, proved to provide little additional diagnostic information in the multivariate models, and, hence, are not included in the score. This applies, in particular, to the presence of palmoplantar pustulosis, which was found in 22 of 102 patients with NBO, but not in the other diseases. Also not included in the multivariate models were the presence of hyperostosis and fractured vertebrae. Although hyperostosis was a quite reliable indicator of NBO, it appeared in only a small fraction of NBO cases, limiting its use in a diagnostic scoring method. Hyperostosis can occasionally be seen in chronic bacterial osteomyelitis and in patients in a good state of health who have benign bone processes such as osteoid osteoma or osteochondroma. Vertebral involvement in NBO (observed in 37 of 102 patients) seemed to be a valuable parameter, but this also occurred in malignancy (5 of 48 patients) and bacterial osteomyelitis (1 of 22 patients). Fractured vertebrae are less indicative of NBO than hyperostosis and, more importantly, also appear in patients with malignant bone lesions (and, even if not seen in this collective cohort, possibly in those with benign bone lesions). Therefore, and because such fractures are often associated with additional complications, their isolated presence may warrant comprehensive diagnostic measures. Clavicular lesions were observed only in patients with NBO in this collective data set and seemed to be more specific. In particular, symmetric bone lesions occurred only in those with NBO, except in 2 patients with benign bone tumors. For a clinician, the number of bone lesions is an important diagnostic tool in identifying NBO. Our results showed that the number of radiologically proven bone lesions, including those on conventional radiographs, bone scans, MRI, and computed tomography, was significantly higher in patients with NBO. Bone scan should show an uptake in affected bone lesions in patients with NBO, those with bacterial

7 1158 JANSSON ET AL osteomyelitis, and those with most other malignancies. Because a bone scan is of limited value in benign bone tumors as well as in bacterial osteomyelitis, this procedure was not carried out in all patients, and therefore this parameter was not evaluable (79 missing values). Histologic results might be a helpful parameter for diagnosing NBO, even if the findings are nonspecific (6,24). Whereas, typically, histology can be used to reveal inflammation, fibrosis, or sclerosis, it can also be used to exclude malignancy or to detect an organism. In our collective data set, histologic investigations were not performed in many patients with NBO, benign bone tumors, or bacterial osteomyelitis. Therefore, the clinical score is applicable for diagnosing NBO in the absence of histologic results. Usage of the score might ease the decision as to whether a biopsy is indicated. Our proposed clinical procedure is shown in Figure 1, assuming that the initial or dominant symptom is bone pain in an otherwise-healthy patient. If a localized-joint radiograph reveals a bone lesion that does not show uptake on the whole-body bone scan, then a biopsy should be performed. When the radiologically proven lesion shows uptake on the scan but no more lesions can be detected, the score might help to decide whether a biopsy is indicated. If more than 1 lesion shows uptake on the scan, then biopsy can be postponed. In the case of scores below 29, the patient should be monitored carefully. When the duration of bone pain is less than 3 weeks, conventional radiography does not necessarily always show a lesion. If the local symptoms are severe and the radiograph is without pathologic findings, a local MRI might help to refine the diagnosis. If MRI shows suspected lesions, a bone scan should be performed (as outlined above). Even though palmoplantar pustulosis could not be used in a regression-derived score, it nevertheless seems to be very specific and could be valid in the diagnosis of NBO in patients with intermediate scores from 29 to 38. Diagnosing NBO by exclusion is a challenge for the clinician. Due to uncertainty, invasive and possibly unnecessary diagnostic and therapeutic procedures are often carried out. Nevertheless, in some cases, additional procedures might be necessary in order to ensure that a more serious disease is not overlooked. The clinical score proposed herein may be helpful in making this decision. However, prospective evaluations of independent patient populations will be required to ascertain the properties of this diagnostic tool. ACKNOWLEDGMENTS We thank Roland Duerr, MD, Department of Orthopedic Surgery, for assistance in patient recruitment, Karl Schneider, MD, and Birgit Kammer, MD, Department of Pediatrics, for radiologic assistance, and Veit Grote, MD, Department of Pediatrics, for assistance with development of the manuscript. AUTHOR CONTRIBUTIONS Dr. V. Jansson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study design. A. F. Jansson, Müller, Belohradsky, V. Jansson. Acquisition of data. A. F. Jansson, Gliera, Wintergerst, Belohradsky, V. Jansson. Analysis and interpretation of data. A. F. Jansson, Müller, Gliera, Ankerst, Wintergerst, Belohradsky, V. Jansson. Manuscript preparation. A. F. Jansson, Müller, Gliera, Ankerst, Wintergerst, Belohradsky, V. Jansson. Statistical analysis. Müller, Ankerst. REFERENCES 1. Hayem G, Bouchaud-Chabot A, Benali K, Roux S, Palazzo E, Silbermann-Hoffman O, et al. SAPHO syndrome: a long-term follow-up study of 120 cases. Semin Arthritis Rheum 1999;29: Benhamou CL, Chamot AM, Kahn MF. Synovitis-acne-pustulosis hyperostosis-osteomyelitis syndrome (SAPHO): a new syndrome among the spondyloarthropathies? Clin Exp Rheumatol 1988;6: Beretta-Piccoli BC, Sauvain MJ, Gal I, Schibler A, Saurenmann T, Kressebuch H, et al. Synovitis, acne, pustulosis, hyperostosis, osteitis (SAPHO) syndrome in childhood: a report of ten cases and review of the literature. Eur J Pediatr 2000;159: Kahn MF. Current status of the SAPHO syndrome. Presse Med 1995;24: In French. 5. Bianchi G, Marinelli A, Frizziero A, Mercuri M. Hyperostosis and osteitis in Sapho syndrome: conservative or surgical treatment. Chir Organi Mov 2004;89: Bjorksten B, Boquist L. Histopathological aspects of chronic recurrent multifocal osteomyelitis. J Bone Joint Surg Br 1980;62: Carr AJ, Cole WG, Roberton DM, Chow CW. Chronic multifocal osteomyelitis. J Bone Joint Surg Br 1993;75: El-Shanti HI, Ferguson PJ. Chronic recurrent multifocal osteomyelitis: a concise review and genetic update. Clin Orthop Relat Res 2007;462: Girschick HJ, Raab P, Surbaum S, Trusen A, Kirschner S, Schneider P, et al. Chronic non-bacterial osteomyelitis in children. Ann Rheum Dis 2005;64: Golla A, Jansson A, Ramser J, Hellebrand H, Zahn R, Meitinger T, et al. Chronic recurrent multifocal osteomyelitis (CRMO): evidence for a susceptibility gene located on chromosome 18q q22. Eur J Hum Genet 2002;10: Huber AM, Lam PY, Duffy CM, Yeung RS, Ditchfield M, Laxer D, et al. Chronic recurrent multifocal osteomyelitis: clinical outcomes after more than five years of follow-up. J Pediatr 2002;141: Job-Deslandre C, Krebs S, Kahan A. Chronic recurrent multifocal osteomyelitis: five-year outcomes in 14 pediatric cases. Joint Bone Spine 2001;68: Jansson A, Renner ED, Ramser J, Mayer A, Haban M, Meindl A,

8 CLINICAL SCORE FOR NONBACTERIAL OSTEITIS 1159 et al. Classification of non-bacterial osteitis: retrospective study of clinical, immunological and genetic aspects in 89 patients. Rheumatology (Oxford) 2007;46: Jurik AG. Chronic recurrent multifocal osteomyelitis. Semin Musculoskelet Radiol 2004;8: Hyodoh K, Sugimoto H. Pustulotic arthro-osteitis: defining the radiologic spectrum of the disease. Semin Musculoskelet Radiol 2001;5: Jansson A, Borte M, Boschow G, Dressler F, Gruubner I, Kekow M, et al, for the German Society for Paediatric and Adolescent Rheumatology. Non-bacterial osteitis in children and adults: consensus statement of the 8th Worlitz Expert Meeting Monatsschrift Kinderheilkunde 2006;154: Sing T, Sander O, Beerenwinkel N, Lengauer T. ROCR: visualizing classifier performance in R. Bioinformatics 2005;21: Offiah AC. Acute osteomyelitis, septic arthritis and discitis: differences between neonates and older children. Eur J Radiol 2006;60: Sundaram M, McDonald D, Engel E, Rotman M, Siegfried EC. Chronic recurrent multifocal osteomyelitis: an evolving clinical and radiological spectrum. Skeletal Radiol 1996;25: Ferroni A. Epidemiology and bacteriological diagnosis of paediatric acute osteoarticular infections. Arch Pediatr 2007;14 Suppl 2:S91 6. In French. 21. Chometon S, Benito Y, Chaker M, Boisset S, Ploton C, Berard J, et al. Specific real-time polymerase chain reaction places Kingella kingae as the most common cause of osteoarticular infections in young children. Pediatr Infect Dis J 2007;26: Rasool MN. Osseous manifestations of tuberculosis in children. J Pediatr Orthop 2001;21: Cerase A, Priolo F. Skeletal benign bone-forming lesions. Eur J Radiol 1998;27 Suppl 1:S Girschick HJ, Huppertz HI, Harmsen D, Krauspe R, Muller- Hermelink HK, Papadopoulos T. Chronic recurrent multifocal osteomyelitis in children: diagnostic value of histopathology and microbial testing. Hum Pathol 1999;30:59 65.