This study has been approved by the local ethics committee (No. 36-070 ex 23/24). It was based on a retrospective analysis of 44.762 patients with an MRI of the knee and 21.550 patients with an MRI of the shoulder taken between 01.01.2007 and 01.03.2020 at a single radiology centre. Further details on the patient cohorts have been published previously [15, 16].

As depicted in Fig. 1, 504 patients with cartilage tumours of the knee and 64 patients with EC/ACT of the shoulder had to be excluded from the current study, as they had only received one MRI (baseline MRI) throughout the study period. The remaining 176 patients (147 patients with EC/ACT of the knee and 29 patients with EC/ACT of the shoulder) had undergone baseline MRI and at least one follow-up MRI. As 6 of these 176 patients had two cartilaginous tumours at the same time, 182 lesions in total (171 ECs, 11 ACTs) could be finally included. Patients had been referred to the radiology institute for different reasons: 38 patients (21.6%) had received MRI due to suspected/known cartilage lesion, 97 patients (55.1%) due to other reasons, and indication of MRI had not been documented in 41 cases (23.3%).

Flow-chart depicting the filtering of patients with an EC/ACT of the knee or shoulder, who received FU-MRIs

In a first step, baseline and follow-up MRI reports (written by expert radiologists) of the included 176 patients were analysed by one of the co-authors, who had undergone extensive training regarding image interpretation of benign and intermediate cartilaginous tumours. All reports were re-examined together with the corresponding MRI scans and then filtered for the following data: patient gender and age, date and number of MRIs performed, tumour size, periosteal reaction, medullary oedema, endosteal scalloping, affected bone, tumour location within the affected bone, and location in relation to medullary canal.

Second, an expert orthopaedic oncologist re-examined all baseline and follow-up MRIs of the eleven included ACTs, again focusing on the above-mentioned tumour characteristics. All the findings of the radiologists were re-evaluated by this orthopaedic oncologist, who reached similar results.

Intraosseous tumours presenting as smooth or lobulated lesions with tiny calcifications and showing high signal intensity on PD-FS-weighted (proton density, fat-suppressed) images and low signal intensity on PD-weighted as well as T1-weighted images were referred to as cartilage lesions.

Radiological differentiation between ECs and ACTs was performed based on the following criteria: Lesion size over 4.9 cm, deep endosteal scalloping (involvement of more than 2/3 of the bone cortex), tumour-related medullary oedema, and periosteal reaction. These features are highly indicative of aggressive tumour behaviour [4, 8, 9, 17]. Therefore, all lesions positive for at least one of these features were classified as ACTs.

Determination of tumour size was performed via measurement of the maximum lesion diameter in the coronal view. For size measurement, anatomical structures adjacent to the lesions were utilised as supporting landmarks to visualise the tumour in the same sectional plane across baseline and all follow-up MRIs.

To evaluate the accuracy and reproducibility of size determination, measurements that had been carried out at primary time point of data acquisition (measurement 1) were repeated one year thereafter (measurement 2) by an extensively trained co-author. Estimation of intraindividual reproducibility regarding tumour size evaluation at baseline and last follow-up was performed via calculation of coefficients of variation (ratio of standard deviation of measurement 1 and 2 to the mean of measurement 1 and 2 followed by multiplication with 100%). Comparison of measurements 1 and 2 for cartilage tumours of the knee revealed mean coefficients of variation of 1.8% (lesion size at baseline) and 1.8% (lesion size at last follow-up), whereas for shoulder tumours, mean coefficients of variation were 1.7% (lesion size at baseline) and 2.0% (lesion size at last follow-up). Furthermore, a comparison of tumour growth rates based on measurements 1 and 2 was performed via Bland-Altman plots as depicted by Fig. 2. As all of these evaluations indicated high reproducibility of tumour size measurement, changes in lesion size of at least 1 mm were referred to as statistically significant.

Comparison of tumour growth rates of knee and shoulder as calculated on the basis of measurements 1 and 2. For calculation of the difference between measures, measurement 2 was subtracted from measurement 1

After having completed radiological and clinical data acquisition, all patients with radiologically diagnosed ACTs (n = 11) were searched for in the hospital’s medical documentation system to investigate whether these patients had undergone tumour surgery/biopsy with histological tumour evaluation. However, only one of eleven patients had undergone tumour surgery and histopathology confirmed the radiological diagnosis of an ACT.

Patients received MRI examinations via 3-T MRI systems with a 15/18 channel coil for knee and a 16 channel coil for shoulder (Siemens Magnetom Skyra/Siemens Magnetom Vida). Detailed descriptions of all sequences acquired have been published previously [15, 16].

All statistical analyses were performed with R version 4.2.1 (© 2022 The R Foundation for Statistical Computing). Normally distributed variables are represented as means with corresponding standard deviations, and non-normally distributed variables are given as medians with corresponding interquartile ranges (IQR). Fisher’s exact test was applied for the assessment of statistical differences in ordinary variables and t-test/Wilcoxon rank sum test was used for normally distributed/non-normally distributed metric variables. A p-value < 0.05 was considered statistically significant.

The following parameters were calculated for each lesion based on the data obtained: follow-up time, alteration of tumour size, tumour growth, tumour growth rate, alteration of periosteal reaction, alteration of medullary oedema and alteration of endosteal scalloping.