Our investigation describes in detail the imaging findings expected in vertebral and mandibular lesions of CNO and their evolution during the course of the disease.

The demographic characteristics of the patients in this study do not differ much from other published studies. More females than males were affected, with symptom onset usually during early adolescence. Our time until diagnosis was somewhat longer than other studies, which may reflect still some lack of awareness of this entity [2, 3, 17, 24]. Back pain was a significant clinical marker of vertebral disease, as already previously reported [17, 24]. The total number of vertebral bone lesions per patient was also similar to a previous report [24].

Follow-up intervals were quite variable between our patients since our center did not follow a specific rule to time out follow-up exams, and rather based the execution of imaging exams on clinical aggravation.

In patients with spine lesions, involvement of multiple vertebral levels was the most common presentation, with dorsal levels predominantly affected. Vertebral body height loss was an important complication, showing up in more than half of our patients and leading to spinal curvature deformities in some cases. Resolution of vertebral inflammatory lesions, total or partial, occurred in all patients, but recurrence appeared in about a third. Sacral lesions, when present, always affected the sacroiliac joints.

Several studies showed that vertebral involvement is frequent [3, 11, 12, 24,25,26], ranging up to 38% in the general population [12] and to 81% in a pediatric population [26], suggesting vertebral involvement might be higher in children. In our sample, 45% of the patients with vertebral lesions had no other anatomical locations affected, suggesting that limited involvement of the spine in CNO might also be frequent. Similar to other studies, spine lesions most frequently affected the thoracic spine [12, 24, 26]. In our study, 82% of the patients had multiple vertebral levels affected. Another study described multifocal vertebral involvement in 60% of patients in a pediatric population [26]. These results suggest multifocality as a common finding of spine CNO lesions in children.

Most of our patients (55%) had lesions with vertebral height loss. A previous large study described vertebral fractures in 17.5% of the patients with spine involvement [2]. Considering that fractures usually result in vertebral height loss, this value is low when compared to ours. However, our results are more similar to those of other studies where spinal lesions were specifically evaluated with MRI, which found vertebral body deformities in 52% and vertebral height loss in 64% of the patients with spine involvement [24, 26]. The patients in our study all performed MRI, which is more sensitive than other imaging modalities and may explain these different results. We also showed that vertebral lesions in children can lead to abnormal kyphosis and scoliosis, with 36% of our patients having spinal deformities, similar to what has been previously reported [12, 26].

Vertebral endplate involvement and disc lesions occurred in 73% and 55% of our patients, respectively, differing from a study by Guariento et al., where disc involvement was found in 26% and endplate abnormality in 57% [26], but these values show that endplate and disc involvement may be frequent regardless. In our study, these abnormalities almost always occurred adjacent to vertebrae with platyspondyly, a tendency that Guariento et al. also recognized. Finally, although the vertebral body is the region that is classically injured in CNO, with possible extension to the pedicles, we had one patient who showed isolated involvement of the posterior vertebral elements in several levels. This pattern was also described in one patient by Guariento et al., highlighting that, although rare, it should be recognized.

Active inflammatory changes on imaging appeared in almost all our patients, except two, in which only sequelae of previous vertebral lesions were found. All patients with active inflammation had bone edema, and almost half had surrounding soft tissue inflammation with a para-vertebral or pre-sacral distribution. These findings on MRI are not specific of CNO and rather represent an inflammatory process, but are similar to other descriptions of CNO lesions, vertebral or otherwise [17, 27, 28].

Total or partial resolution of vertebral inflammatory lesions on follow-up imaging was present in all our patients. Treatment approaches across our patients were variable, and due to a small sample size we could not establish a relationship between the treatment used and the degree of lesion resolution. However, all our patients were treated, and our results acknowledge the response of vertebral lesions to therapy, but the spontaneous resolution of spine lesions could not be evaluated. The best treatment strategy for CNO is yet to be completely determined. NSAIDs have been shown to be an effective first treatment option [29]. Consensus treatment plans for CNO refractory to NSAIDs have been established [30], but the comparison of effectiveness between different treatments is still lacking [31]. Other studies proposed bisphosphonates as an effective treatment option for CNO patients with vertebral lesions, with good results [21, 24]. One study demonstrated partial or complete resolution of vertebral MRI hyperintensities in every patient treated with pamidronate [24]. This emphasizes how imaging may be an important mean to monitor the evolution of lesions and treatment response. As such, whole-body MRI has been proposed as the best imaging modality to monitor disease activity, but the ideal follow-up intervals have not been clearly established [28, 29, 32].

Recurrence rates of CNO have been variably reported, from as low as 16% to as high as 83%, and may vary according to follow-up time and treatment used [13, 21, 33, 34]. In our study, the recurrence rate of vertebral lesions on MRI was inside this interval, but recurrence rates on imaging may be lower than clinical recurrence of symptoms, since some of our patients had recurrent symptoms not always accompanied by imaging findings. We also found that all patients with recurrent vertebral lesions on MRI had its recurrence at vertebral levels not previously affected, and that recurrence can appear as late as after 4 years of follow-up.

Reported sacral involvement is about 19% in CNO patients with vertebral lesions [24, 26], which is similar to what we found, but varies from 24 to 72% in all CNO patients [17, 35]. Sacroiliitis has been reported in 72% of CNO patients, with most having bilateral involvement, similar to our results [35]. A previous study showed that unilateral sacroiliac involvement is more frequent when only clinical evaluation is performed, but bilateral involvement is more frequent when imaging is used [36]. This again underlies the importance of sensitive imaging studies in the investigation of CNO. d’Angelo et al. described sacroiliac involvement as affecting one (iliac or sacral) or both sides of the sacroiliac joint, or the sacroiliac synovial space [17], findings that have also been observed in our study.

CNO is a diagnosis of exclusion, and its imaging appearance can be non-specific, so the differential diagnosis of vertebral lesions and platyspondyly in the pediatric age should always be considered. Infectious osteomyelitis, which can be very hard to differentiate from CNO solely on imaging, should be excluded based on clinical features and other complementary diagnostic tests, sometimes requiring biopsy. Depending on the epidemiologic context, tuberculous and other granulomatous infectious diseases should be suspected. Langerhans cell histiocytosis is the most common cause of vertebra plana (loss of almost the entire vertebral body height) in children and can have soft tissue involvement, underscoring its inclusion in the differential. However, Langerhans cell histiocytosis may have other differentiating features on imaging that are not typical of CNO, and biopsy can be diagnostic. Lymphoproliferative disorders can infiltrate vertebral bones and sometimes resemble osteomyelitis, but, unlike CNO lesions, tend to progress along adjacent vertebral levels with a more infiltrative pattern. Bone metastasis can present heterogeneously, possibly resembling osteomyelitis, but are rare in this age group and should be considered mainly if there is a known primary tumor.

Our study demonstrated that the mandible can be a site of unifocal disease, typically affecting its ramus, with prominent bony changes and soft tissue inflammation. These results mirror those of the current literature. The mandible is recognized as a classic site of involvement in CNO [2, 13, 17, 28] and is a known site of unifocal disease [13]. Chronic osteomyelitis usually affects one hemimandible, involving the ascending ramus and the condylar process [37, 38]. Prevalence of mandibular lesions varies from 2 to 21% [2, 13].

MRI is sensitive for early detection of bone and adjacent soft tissue inflammation in patients with mandibular osteomyelitis [37], but CT can also be useful since these patients usually have extensive bone changes [38, 39]. Previous reports describe these lesions as having mixed lytic and sclerotic abnormalities, diffusely distributed, with areas of resorption of medullary and cortical bone, sequestrum, and periosteal reaction. Inflammation of surrounding soft tissues is a common feature, best seen on MRI [37, 39]. The patients in our study mimic these descriptions.

Another important point is how these lesions evolve over time. When active inflammation subsides, the bone matrix becomes more sclerotic, with less pronounced lytic changes, with redefinition of bone contours, and reduction of contrast uptake; periosteal new bone apposition during acute phases may cause persistence of bone thickening, leading to mandibular enlargement [37]. This is a pattern we observed in our patients. Even though active inflammation improved, sequelae from the lesions persisted.

Imaging helps evaluate response to treatment. Bisphosphonates are described as a treatment option for mandibular chronic inflammatory lesions [38, 40, 41], but we could not evaluate the impact of treatment in mandibular lesions due to our small sample size.

The main limitations of this study are its small sample size and its retrospective nature, which hindered a more reliable evaluation of lesion patterns and relationships, and limited the generalizability of the results, possibly concealing rarer manifestations not included in our results. We cannot exclude selection biases as the patients treated in our pediatric referral hospital were probably referred due to severe manifestations, excluding milder cases from our analysis. Also, our clinicians based the execution of follow-up imaging on clinical aggravation, meaning that patients with milder clinical episodes may not have been imaged. A case–control prospective study could better evaluate the relationship between imaging and clinical features, and larger sample sizes and follow-up times could unravel features and imaging patterns not observed in our study.

In CNO, vertebral involvement is not rare, and neuroradiologists may encounter this disease during their work life, highlighting the importance of knowing it.

Vertebral lesions are often multiple, affecting predominantly dorsal levels, with most patients showing vertebral body height loss. Resolution of vertebral lesions, partial or total, is frequent, but recurrence is not uncommon. Sacral lesions are a feature of CNO, frequently affecting the sacroiliac joints. Mandibular lesions may be a site of unifocal disease, and the mandibular ramus is typically affected, with surrounding soft tissue inflammation.

For practicing neuroradiologists, a key feature for distinguishing CNO lesions is the involvement of multiple anatomical sites, especially if typical sites are involved, like the long bone metaphysis of the lower extremities, pelvis, clavicles, and spine. For isolated involvement of the spine, involvement of several levels with diffuse edema of the vertebral body, platyspondyly, and predilection for dorsal levels should raise suspicion. Bilateral involvement of the sacroiliac joints and adjacent bone also seems to be characteristic of CNO. Isolated mandible lesions represent a more challenging diagnosis, but lesions that only partially recover over time and recurrence at the same site may be suggestive.

Although the prognosis of CNO is considered good, with the resolution of symptoms in most patients, about a quarter can have persistent disease and sequelae [42, 43]. Studies with longer periods of follow-up may be necessary to investigate how CNO evolves over time and the factors that influence its progression. The ideal follow-up intervals for radiological monitoring of CNO patients are still not determined, and future longitudinal studies on the long-term outcomes of these patients could help clarify this question. Validating the imaging findings in a larger, more diverse population could also improve the generalizability of our results. Finally, exploring the underlying pathophysiological mechanisms leading to specific imaging findings could provide directions for future research.