The current study was based on shoulder MRI reports of 21.550 patients who underwent imaging at a single radiology institute throughout a time span of 13.2 years. A prevalence of 0.43% for cartilaginous lesions around the shoulder joint was revealed.

As cartilaginous lesions of the shoulder joint are typically asymptomatic, their presence is mostly revealed accidentally via MRIs made for different reasons, making it hard to estimate the true prevalence of enchondromas and ACTs.

To the best of the authors’ knowledge, prevalence and characteristics of cartilage tumours in the shoulder girdle have only been examined in one small-sized study so far [13]: Hong et al. analysed 477 patients with MRI scans and reached a figure of 2.1% for the prevalence of cartilage lesions around the shoulder joint. The herein described study—based on a 45 times larger cohort (n = 21.550) – found a noticeably lower prevalence of 0.43%.

While Hong et al. reported on 10 patients with an enchondroma and no patient with an ACT at all [13], our study cohort included 93 patients with altogether 89 enchondromas (91.8%) and 8 suspected ACTs (8.2%). Only 8 lesions in our cohort were classified as intermediate concerning their dignity, resulting in an overall prevalence of 0.04% for ACTs in the shoulder joint. As – to the best of the authors’ knowledge – no previous study has ever focused on the prevalence of ACTs of the shoulder joint, this is the first figure elucidating this issue.

Two different research groups have shown that the frequency of chondrosarcomas diagnosed – and especially the one of ACTs – has increased over the last decades, most likely due to intensified MRI screening, causing a rising number of accidental findings [4, 15]. Our study has not been able to validate these results for ACTs around the shoulder joint since only 8 patients presented with an ACT like lesion, too few to identify a trend (Fig. 3).

MRI scan of an ACT (maximal diameter of 5.2 cm; deep endosteal scalloping; medullary oedema) of the left humerus: A proton density, blade with fat suppression, coronal; B proton density + t2, fat suppression, dixon turbo spin echo, sagittal; C proton density, blade with fat suppression, transversal; D t1, coronal, turbo spin echo. The arrows point at endosteal scalloping, whereas the asterisk highlights medullary edema

Furthermore, Davies et al. showed that 85% more patients with an enchondroma were referred to an UK-based specialist orthopaedic oncology unit between 2009–2018 than in the time span of 1999–2008, indicating a rise in incidence of enchondromas over the last decades [4]. Notably, the current study, performed between 2007 and 2020, could not confirm an increase over this time span as far as the prevalence of enchondromas of the shoulder joint is concerned.

Analysis of tumour characteristics showed that ACTs (mean size 5.5 cm) were significantly (p < 0.001) larger than enchondromas (mean size 2.0 cm), thereby confirming previous results by Kendell et al. who analysed intraosseous cartilaginous lesions of the fibula and highlighted the importance of lesion size as a marker of differentiation between benign and intermediate cartilaginous tumours. They found that lesions smaller than 4 cm tended to be enchondromas, whereas a size of more than 4 cm was indicative of an atypical cartilaginous tumour (formerly known as low-grade chondrosarcoma) [16].

The majority of enchondromas of the shoulder presented were located in the proximal humerus (96.6%) (Fig. 4), whereas only three ECs were found in the scapula (3.4%). In literature, the scapula is referred to as an unusual localization of enchondroma with only two retrospective studies [13, 17] and a few case reports [18,19,20] having published the finding of an EC at this specific anatomical site (to the best of the authors’ knowledge, altogether eight ECs of the scapula have been reported in literature up to now). Song et al. have only recently evaluated tumours of the scapula and found that among 108 cases of benign and malignant lesions of the scapula, four patients (3.7%) presented with an enchondroma [17] indicating that enchondroma represents a rare but relevant differential diagnosis when dealing with tumours of the scapula. Hong et al. discovered that 10% of shoulder enchondromas reside in the scapula. Authors of that study, however, were only able to analyse tumour characteristics of 10 cartilaginous shoulder lesions with one of them being located in the scapula [13]. As the current study was based on 97 ECs/ACTs of the shoulder joint and implicates a lower frequency for scapular ECs (3.4%), it can be assumed that the figure of Hong et al. on this matter (10%) is overestimated, as their study was based on a significantly smaller cohort [13].

MRI scan of an enchondroma of the right humerus: A proton density, blade with fat suppression, coronal; B proton density + t2, sagittal; C proton density, blade with fat suppression, transversal; D t1, coronal, spin echo

There was a significant difference between ECs and ACTs (p = 0.009) regarding location in relation to medullary canal, with ACTs showing eccentric location significantly more often (100%) than ECs (52.8%). While in a cohort by Bui et al., reporting on 11 patients with eccentric ECs of long bones all tumours (100%) all were associated with cortical scalloping [21], the current data cannot replicate these results, as only 2 of 47 (4.3%) eccentrically located ECs presented with superficial scalloping. A major difference between the studies is represented by the fact that Bui et al. only examined eccentric lesions of long bones around the knee joint (10 tumours of the femur and 1 of the tibia) [21].

Regarding location within long bones, ECs represent intramedullary lesions typically residing in the metaphysis of long bones [10]. The theory that ECs arise out of physeal rests, which were trapped in the metaphysis of growing long bones, is mainly based on this finding [1]. However, this widely believed theory has been challenged by Douis et al. [22], who did not find any displaced cartilage in the metaphysis of long bones in skeletally immature individuals. However, ECs cannot solely be found in the metaphysis and diaphysis, with Potter et al. [23] reporting that 6.7% of patients with an EC affecting long bones (33 out of 508 patients) showed a benign cartilaginous tumour originating from the epiphysis, with the majority of these lesions residing in the proximal humerus (30%). Hong et al. discovered in their tumour cohort that even more ECs of the proximal humerus were located in the epiphysis (55.6%) than in the metaphysis (44.4%) [13]. This finding cannot be supported by our data, suggesting that only a minority of ECs (6.7%) are located in the epiphysis, thereby confirming the finding of Potter et al. for the frequency of epiphyseal location in benign intraosseous tumours of long bones [23].

Some limitations have to be considered when interpreting the results: First, diagnosis of all cartilaginous lesions was MRI-based only. Radiological criteria to differentiate enchondromas from atypical chondromatous tumours are not utterly defined, though. Second, due to the benign nature of ECs, the vast majority of patients did not undergo biopsy, wherefore histopathological assessment was not part of the diagnostic approach. Considering that histological differentiation between EC and ACT depends on the quality of the biopsy and may be prone to sampling error, it can be assumed that additional histological examination would not have significantly altered results obtained [24]. Two further studies have confirmed that histological differentiation of ECs and ACTs represents a difficult task for pathologists, as high interobserver variability in diagnosis as well as grading of cartilaginous lesions could be detected [25, 26]. Flemming et al. in 2000 and Eefting et al. in 2007 proposed a diagnostic approach that combines radiological and histological assessment for differentiation of ECs and ACTs, as pathohistological examination solely leads to unreliable diagnoses [1, 26]. Miwa et al., however, discovered that comprehensive assessment of radiological examinations solely shows high accuracy for evaluation of dignity in chondromatous lesions [27]. Furthermore, bioptic validation of conclusive radiological diagnosis of a benign intraosseous chondromatous lesion represents an unnecessary burden for a patient, which is why the majority of ECs do not undergo histological examination. Therefore, it is hard to identify a cohort of patients with histopathologically confirmed ECs, and it appears that authors of an epidemiological study with the ambition to find the prevalence of ECs and ACTs have to accept that such a study either must be based on a small cohort, where all tumours can realistically be histologically confirmed or they must accept imaging-derived diagnosis of lesions with the major advantage that a high patient number can be included. As imaging-based diagnosis has proven to do very well, even in distinguishing ECs from ACTs [27], and as significant epidemiological studies subsist on patient cohorts as large as possible, it can be assumed that a study with the ambition of finding the true prevalence of ECs and ACTs has to do without histology.

Due to the criteria chosen to differentiate between EC and ACT, statistically significant results with regards to features as tumour size emerged (Table 1). Yet, also literature [6, 8, 14] confirms that these features vary between the two cartilaginous neoplasms. Another limitation of this study is presented by the retrospective study design, as MRIs had usually been performed for other shoulder pathologies, with the incidental finding of cartilage tumours.

In summary, the herein presented large-sized study revealed an overall prevalence of 0.43% for benign and intermediate chondromatous lesions around the shoulder joint, with 8.2% of these lesions exhibiting at least one MRI feature highly suspicious of ACT. Prevalence of ECs around the shoulder joint has probably been slightly overestimated in the past, however, as 1 in 233 patients will show an enchondroma in an MRI of the shoulder, it remains an important differential diagnosis not to be mistaken for other pathologic entities.