Standardized reporting and quantification of whole-body MRI findings in children with chronic non-bacterial osteomyelitis treated with pamidronate

Chronic non-bacterial osteomyelitis (CNO) is a skeletal disorder in children and adolescents. CNO is characterized by sterile and often multifocal inflammatory bone lesions [1] with a predilection for the metaphyses of the long tubular bones in the lower extremities, the axial skeleton and the clavicles. During the disease course bone lesions can vanish or persist and bone lesions may occur at new sites [2‐6]. Children with CNO present clinically with musculoskeletal complaints such as pain and swelling at the skeletal sites involved and spinal lesions can lead to vertebral fractures and risk of kyphoscoliosis [7, 8]. However, radiologically active lesions may be asymptomatic [2, 3, 9].

In CNO, MRI is useful to assess active and chronic bone inflammation, synovitis and adjacent soft-tissue edema. Whole-body MRI (WBMRI) is increasingly being used initially at time of diagnosis to identify clinical silent lesions and to detect multifocal bone lesions at typical locations. Follow-up scans are used to detect complications, new lesions and clinically silent lesions when monitoring the disease during treatment [2, 5, 10‐14]. Due to rather variable CNO disease courses with vanishing active lesions and potential occurrence of new active or chronic lesions a global assessment and reporting of disease activity is important. The purpose of developing a standardize radiological outcome measure in CNO is to develop a tool that can be used to describe changes in radiological disease activity both in clinical studies and to monitor disease activity in a daily clinical setting.

A radiological index for non-bacterial osteitis (RINBO) has been suggested for standardized reporting of WBMRI in children with CNO [2]. RINBO includes four parameters of interest: number and maximal size of radiological active lesions, extramedullary changes and spine involvement encompassing both active and chronic lesions in the form of osseous hyperostosis and vertebral deformity (Table 1). A RINBO scoring is easy to perform and results in an individual summary patient score describing radiological disease activity. However, it has not been evaluated in longitudinal studies.

Treatment of CNO aims to reduce inflammation, osteolytic progression, disability and pain. The therapeutic approach to CNO is based on low level of evidence. Several treatment algorithms have been suggested [6, 15]. Non-steroidal anti-inflammatory drugs (NSAIDs) are recommended as first-line therapy [3, 16]. Children with multiple lesions or spinal involvement are considered to have more challenging disease courses [6, 17], necessitating treatment with second-line pharmaceuticals [6, 18]. In cases-series, the amino bisphosphonate pamidronate has been reported to be effective in improving clinical and radiological disease activity [6, 8, 19‐29] while other studies have suggested treatment with biological disease modifying anti-rheumatic drugs (bDMARDS). In our institution children with symmetric multifocal or spinal bone inflammation and a poor clinical response to NSAID´s were offered second-line treatment with pamidronate [6].

The objective of this retrospective study was to test a modified radiological index for non-bacterial osteitis (mRINBO) for detecting changes in radiological disease activity by WBMRI and to assess the therapeutic effect of pamidronate in children with symmetrical multifocal or spinal bone inflammation treated with pamidronate.

Modifications of the original radiological index for non-bacterial osteitis (RINBO) [35] regarding the size of RALmax and exclusion of RAL in the metatarsal bones were implemented, and the modified mRINBO was used for standardized reporting and quantification of WBMRI findings in children with CNO to evaluate the effect of pamidronate therapy. mRINBO was assessed initially and after one year of pamidronate therapy.

We observed an improvement in radiological disease activity assessed by mRINBO following one-year therapy. From baseline to year-one the mRINBO decreased from a median of 5 [IQR, 4–7] to a median of 4 [IQR, 3–5] (p = 0.05) driven by a significant decrease in the number of RALs per patient, a decrease in the size of RALmax and resolution of extramedullary changes in all but one bone-lesion situated in the tibia. Signs of reduced radiological activity occurred especially in the pelvis, in the long tubular bones of the lower extremities and in the thoracic spine. These findings are in concordance with prior studies in CNO, that report similar response to pamidronate, when radiological disease manifestations are assessed by MRI [21, 23, 29, 37].

When assessing the size of RALmax we observed, that the largest RAL at baseline was 65 mm and the median size of RALmax was 39 [IQR, 26–45] mm and at year-one, the largest RAL was 78 mm and the median size of RALmax was 28 [IQR, 20–40] mm. In the original RINBO study the largest RAL was considerable larger being 265 mm (mean 64 ± 55 mm) [2]. The size of RAL depends on disease activity, but also on the size of the bone involved, implying that the age of child and the anatomic location can influence the size of RAL and RALmax. CNO lesions vary considerably in size; the largest lesions often occurring in the pelvis and future studies may have to prove that the maximal size of RAL is an important feature of disease activity. Adjustment using relative measurements of RAL taking the various bone sizes into account may also be considered in future studies [38].

The anatomical distribution of bone lesions was in agreement with previous studies in CNO [4, 29]. Metatarsal bone lesions were omitted because assessment of metatarsal lesions revealed that they were often clinically silent. Radiological CNO findings in the feet have previously been reported in a WBMRI study of 53 children with CNO. This study also showed a mismatch between radiological and clinical findings in the metatarsal bones [5]. Furthermore, metatarsal bone lesions are difficult to assess by coronal WBMRI and difficult to differentiate from non-specific hyperintensities in the feet which may be a normal finding in children [36]. Therefore, if lesions in the forefoot are considered of clinical significance, it is recommended to perform site specific MRI assessment or plain radiographs, which may disclose characteristic CNO changes in the feet, including osteolysis, osteosclerosis and new bone formation [9, 39].

In CNO, WBMRI is often used for diagnosis and may be used as follow-up [12]. An advantage of the mRINBO score is that it follows a systematic quantitative approach and results in a summary patient score. In this study, we assessed changes in radiological disease activity in children with CNO treated with pamidronate for one year. However, mRINBO may be used to assess radiological changes independently of the therapy used. So far, assessment of changes in radiological disease activity in children with CNO has previously predominantly been descriptive, mainly reporting the number of bone lesions and changes in bone marrow edema based on expert opinion [3, 8, 21, 23, 24, 30, 37, 40‐44]. Panwar et al. report changes in radiological disease activity in predefined anatomical location based on a WBMRI scoring system. This system encompass the number, relative size and signal intensity of active CNO lesions as well as soft tissue inflammation and structural lesions in the form of osseous hyperostosis and vertebral collapse [29]. Zhao et al. suggest another consensus-driven method for semiquantitative grading of the various CNO lesions by WBMRI, ´chronic nonbacterial osteomyelitis magnetic resonance imaging scoring´ (CROMRIS). This method include both active and structural changes and has an advantage of using a relative size measure of the bone lesion in relation to the bone sizes in affected children [38]. However, these methods do not describe a summary patient score and have to our knowledge not been tested in longitudinal studies, probably because further studies are needed to determine the weighting of each variable.

Monitoring CNO activity is complicated by the fluctuating disease course with spontaneous development of new RALs [3, 39], also during treatment [8, 30]. The mRINBO has the advantage of following a systematic quantitative approach reporting the accumulated radiological disease activity in each individual with CNO. Also, the mRINBO quantification is relatively easy to obtain and takes the important features of CNO into account, considering anatomical locations in addition to lesion number and size. Some anatomical locations may be considered to imply an increased risk such as spinal lesions as they can be osteolytic and lead to vertebral collapse [7, 8, 45]. We observed a statistically significant improvement in the number of spinal RAL, but it was not possible to show improvements in vertebral collapse during one year of pamidronate treatment as has been show in a previous case studies of long lasting pamidronate treatment up to 41 months [8, 19].

Information on clinical disease activity was assessed retrospectively and therefore not acquired in a standardized manner. We were not able to show a correlation between changes in mRINBO and clinical disease activity. A lack of correlation between clinical and radiological disease activity and occurrence of asymptomatic CNO lesions have previously been reported [2, 3]. Our definitions on clinical disease activity were chosen with reference to Wallace et al.’s preliminary criteria for inactive disease and clinical remission in juvenile idiopathic arthritis [31]. Previously, PedCNO has been suggested to describe improvements in disease activity in CNO in prospective studies [3]. PedCNO is a clinical/radiological composite score, that refers to a set of 5 outcome variables in CNO: ESR, number of radiological active bone lesions, severity of disease estimated by the physician, severity of disease estimated by patient or parent, and the Childhood Health Assessment Questionnaire (CHAQ). However, PedCNO has a limitation as the radiological measure only encompass the number of radiological active bone lesions and not lesion size, site, extramedullary changes or chronic changes. In case studies, clinical disease activity has also been assessed as pain on a visual analogue scale [21, 42, 43, 46] and CHAQ [43, 46] and both measures have been shown useful to describe changes in clinical disease activity in CNO.

There were several limitations in our study. First, the size of the study population is limited and the study cohort is non-controlled. It may be possible that the radiological improvement observed can be due to an innate disease course. Secondly, WBMRI was performed on two different MRI scanners and the protocol used did not always include a T1-weighted sequence. Systematically performed T1-weighted sequences might have contributed to a higher sensitivity and specificity in assessing bone lesions than STIR alone. Thirdly, the study population was selected based on inclusion and exclusion criteria regarding CNO severity and uniform imaging regarding modality and timing. Children were treated with pamidronate according to a local treatment algorithm. Six children who received TNF-α inhibitor treatment were excluded from the study as they were mainly treated with bDMARDs due to inflammatory comorbidity and not due to CNO. Exclusion of these children might have led to bias. Also, exclusion of children not fulfilling the imaging criteria might have led to bias as children not examined by WBMRI might have a milder clinical disease course.

In CNO, children receiving pamidronate treatment, radiological disease activity decreased, when assed by mRINBO, indicating a therapeutic effect. The mRINBO score may be a potential method for standardized reporting and quantification of WBMRI findings and radiological changes in children with CNO. However, further development is needed to investigate if mRINBO is a feasible and valid scoring tool. This will include studies of RALmax including considerations using relative size measurements of bone lesions and interrater agreement and reliability analyses in a larger representative cohort. This study also showed, that a standardized WBMRI protocol including coronal STIR and T1-weighted imaging in addition to sagittal imaging of the spine must be recommended to obtain uniform scoring of active and structural radiological changes in CNO.