RPS 616a - Musculoskeletal tumours and body composition quantitation

Purpose:

To meta-analyse the relationship between preoperative sarcopenia and the development of severe postoperative complications in patients undergoing oncological surgery.

Methods and materials:

PubMed and Embase databases were systematically searched from inception until May 2018. Studies reporting on the incidence of severe postoperative complications and radiologically determined preoperative sarcopenia were included. Data was extracted independently by two reviewers. Random effect meta-analyses were applied to estimate the pooled odds ratio (OR) with 95% confidence intervals (95CI) for severe postoperative complications, defined as Clavien-Dindo grade ≥3, including 30-day mortality. Heterogeneity was evaluated with I²-testing. Analyses were performed overall and stratified by the measurement method, tumour location, and publication date.

Results:

A total of 1,924 citations were identified and 53 studies (14,295 patients) included. When measuring the total skeletal muscle area, 43% of the patients were sarcopenic versus 33% when measuring the psoas area. Severe postoperative complications were present in 20% of patients where 30-day mortality was 3%. Preoperative sarcopenia was associated with an increased risk of severe postoperative complications (OR_pooled: 1.44, 95CI: 1.24-16.8, P<0.001, I²=55%) and 30-day mortality (OR_pooled: 2.15, 95CI: 1.46-3.17, P<0.001, I²=14%). A low psoas mass was a stronger predictor for severe postoperative complications compared to a low total skeletal muscle mass (OR_pooled: 2.06, 95CI: 1.37-3.09, OR_pooled: 1.32, 95CI: 1.14-1.53, respectively) and 30-day mortality (OR_pooled: 6.17 (95CI: 2.71-14.08, OR_pooled: 1.80 (95CI: 1.24-2.62), respectively). The effect was independent of tumour location and publication date.

Conclusion:

The presence of low psoas mass prior to surgery as an indicator for sarcopenia is a common phenomenon and is a strong predictor for the development of postoperative complications and 30-days mortality. The presence of low total skeletal muscle mass, which is even more frequent, is a less informative predictor.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Well-differentiated liposarcomas (WDLPS) are difficult to distinguish from lipomas. This distinction is currently made through an invasive biopsy to test for MDM2-amplification. We present a non-invasive alternative using radiomics based on MRI. This work has been accepted for publication by the British Journal of Surgery.

Methods and materials:

Our dataset consisted of 116 patients (58 MDM2-negative lipomas, 58 MDM2-positive WDLPS) with at least a pre-treatment T1-weighted MRI scan who were referred to the Erasmus MC between 2009-2018. When available, T2-weighted scans were included. A clinician manually segmented the tumours, from which 113 radiomics features were extracted. Decision models were created through an automated search amongst a variety of machine learning algorithms to find the combination that maximizes performance. The evaluation was implemented through 100x random-split cross-validation, using 80% for training and model optimisation and 20% for testing. For comparison, the tumours were manually scored by 3 radiologists. Agreement was determined through Cohen’s kappa.

Results:

The T1w-based radiomics model had a mean area under the curve (AUC) of 0.83, a sensitivity of 0.68, and specificity of 0.84. Adding T2w MRI improved the performance to an AUC of 0.89, a sensitivity of 0.74, and a specificity of 0.88. The 3 radiologists had an AUC of 0.74/0.72/0.61, a sensitivity of 0.74/0.91/0.64, and a specificity of 0.55/0.36/0.59. The mean Cohen’s kappa between the radiologists was 0.23, indicating poor interobserver agreement.

Conclusion:

Our radiomics model was able to distinguish WDLPS from lipomas, with a performance superior to 3 experienced radiologists. Hence, radiomics may serve as an objective, non-invasive aid in diagnostic work-up to differentiate between lipomas and WDLPS.

Limitations:

A potential volume bias, which has been assessed in additional subanalyses.

Ethics committee approval

Erasmus MC IRB (MEC-2016-339).

Funding:

NWO #14929-14930, Stichting Coolsingel #567.

Purpose:

The response of desmoid tumours (DTs) to chemotherapy is evaluated with RECIST in daily practice and clinical trials. MRI demonstrates an early change in heterogeneity in responding tumours due to a decrease in cellular area and increase in fibro-necrotic content before a dimensional response. Heterogeneity can be quantified with radiomics. Our aim was to develop radiomics-based response criteria and to compare their performances with the usual criteria.

Methods and materials:

42 patients (median age: 38.2) were included, presenting with progressive DT, MRI at baseline (MRI-0), and early evaluation (3 months later, MRI-1). After signal intensity normalisation, voxel size standardisation, discretisation, and segmentation of DT volume on fat-suppressed contrast-enhanced T1-weighted-imaging, 90 baseline and delta 3D-radiomics features (RFs) were extracted. Using cross-validation and least absolute shrinkage and selection operator penalised Cox regression, a radiomics score was generated. The performances of models based on the radiomics score, modified-RECIST, EASL, Cheson, Choi and revised-Choi criteria from MRI-0 to MRI-1 to predict the progression-free survival (PFS, per RECIST) were assessed with a concordance-index. The results were adjusted for performance-status, tumour volume, prior chemotherapy, current chemotherapy, and beta-catenin mutation.

Results:

There were 10 progressions. The radiomics score included 4 variables. A high score indicated a poor prognosis. The radiomics score correlated with PFS (adjusted hazard ratio=12.02, p=0.002) but none of the usual response criteria. The prognostic model based on the radiomics score had the highest concordance-index (0.85, 95% confidence interval=(0.72-0.99)).

Conclusion:

Quantifying early changes in heterogeneity through a dedicated radiomics score can improve the response evaluation for DT patients undergoing chemotherapy.

Limitations:

A relatively small population study, a small number of events (: progression), and a lack of standardisation of the MRI protocol.

Ethics committee approval

IRB-approved. Written informed consent obtained.

Funding:

No funding was received for this work.

Purpose:

Heterogeneity on DCE-MRIs of sarcomas may be prognostic. The aim was to investigate the best method to extract prognostic data from baseline DCE-MRIs.

Methods and materials:

50 uniformly-treated adults with non-metastatic high-grade sarcomas and pre-treatment DCE-MRIs at 1.5T were included in this retrospective single-centre study. 92 radiomics features (RFs) were extracted at each DCE-MRI phase (11 from t=0s to 88s). Relative changes in RFs (rRFs) since the acquisition baseline were calculated (11x92 rRFs). Curves of rRF as a function of time post-injection were integrated (92 integrated-rRFs [irRFs]). K^trans and the area under the time-intensity curve at 88s parametric maps were computed and 2x92 parametric-RFs (pRFs) were extracted. 5 DCE-MRI-based radiomics models were built on an RFs subset (32s, 64s, 88s), all rRFs, all irRFs, and all pRFs. Two additional models were elaborated as a reference on conventional radiological features and T2-WI RFs. A common machine-learning approach was applied to the radiomics models. Features with p<0.05 at univariate analysis were entered in a LASSO-penalised Cox regression including bootstrapped 10-fold cross-validation. Resulting radiomics scores (RScores) were dichotomised as per their median and entered into multivariate Cox models for predicting metastatic relapse-free survival. Models were compared with a concordance-index.

Results:

Only dichotomised RScores from models based on the rRFs subset, all rRFS, and irRFS correlated with prognostic (p=0.0107-0.0377). The models including all rRFs and irRFs had the highest c-index (0.83) followed by the radiological model. The radiological and full rRFs models were significantly better than the T2-based radiomics model (p=0.02).

Conclusion:

The initial DCE-MRI of STS contains prognostic information. It seems more relevant to make predictions on rRFs instead of pRFs.

Limitations:

A retrospective study, heavy post-processing to homogenise DCE-MRI temporal parameters, with no validation cohort.

Ethics committee approval

IRB approved.

Funding:

No funding was received for this work.

Purpose:

To investigate associations between CT-scan-based body composition (BC) parameters (and their early changes) with progression-free survival (PFS) in metastatic cancer patients treated with immunotherapy.

Methods and materials:

Patients were consecutively included as they were treated with immunotherapy at our institution for a metastatic tumour with an available baseline CT-scan (CT0, ≤28 days before beginning immunotherapy) and early evaluation CT-scan (CT1, 2 months later ± 28 days). At each evaluation, the areas corresponding to psoas alone, skeletal muscle, subcutaneous, visceral, and total adipose tissues at the L3 vertebral level were extracted and weighted by height², providing wSMAI, SMI, SATI, VATI, and TATI, respectively, and their changes (Δt-) from the 1^st day of treatment to CT1. After assessing the optimal cut-point for each BC-parameter (which maximised the bootstrapped concordance-index and balance between the subgroups of patients), correlations with PFS were evaluated using multivariate Cox models.

Results:

Between December 2013-December 2016, 117 patients were included (55 female, median age: 63 years). After cut-point optimisation and adjustment with clinical covariables, 7 BC-parameters correlated with PFS: baseline BMI, SMI, VATI, Δt-wSMAI, Δt-SMI, Δt-SATI, and Δt-VATI (p-values range: <0.001-0.03). At multivariate analysis, 4 remained independently associated with lower PFS, namely: SMI <35.2cm²/m² (HR=3.5, p=0.006), VATI ≥11cm²/m² (HR=2.1, p=0.02), Δt-wSMAI <-0.7 cm²/m²/day (HR=6, p<0.001), and Δt-SATI <-0.02 cm²/m²/day (HR=1.9, p=0.026). Adding these BC-parameters significantly improved the prediction of PFS compared with a model that only included baseline clinical features and BMI (concordance-index=0.76 vs 0.68, p=0.007).

Conclusion:

CT-based BC-parameters and their early changes could help in anticipating the outcome in metastatic cancer patients. Further prospective evaluations are required to validate these findings and to determine whether nutritional interventions are indicated.

Limitations:

The retrospective, single-centre design, manual segmentation, and lack of standardisation in CT acquisition protocols.

Ethics committee approval

IRB approved.

Funding:

No funding was received for this work.

Purpose:

Neurofibromatosis type-1 (NF1) is a dominantly inherited tumour-predisposition syndrome and patients develop large plexiform (PNF) or cutaneous (NF) neurofibromas. High tumour-burden and tumour-growth are indicators for malignant transformation. Studies indicate that large deletions of the NF1 gene and its flanking regions (type-1-deletion) lead to more severe manifestations of NF1 compared to NF1 caused by smaller intragenic changes. Therefore, the purpose of our study was to evaluate the tumour-burden and tumour-growth of patients with type-1-deletions and to compare these with NF1 patients without large deletions of the NF1 gene (non-type-1-deletion patients).

Methods and materials:

We retrospectively evaluated whole-body MRI examinations (1.5T; T1wTSE coronal, T2wTIRM coronal, T2wHASTE TIRM axial, and T2wTSE sagittal) of verified type-1-deletion patients and an age- and sex-matched collective of non-type-1-deletion NF1-patients (both n=38; 20 male vs 18 female; mean age deletion 26.2 years vs non-deletion 25.4 years). Patients were further divided into age-dependent subgroups (0-10; 11-18; 19-49; ≥ 50 years of age). All patients had received follow-up MRIs (mean observed time-period: deletion 6.3 vs non-deletion 5.5 years). Whole-body tumour-volume was semi-automatically assessed (MedX) and tumour-growth over time was calculated.

Results:

NF1 patients with type-1-deletions showed a significantly higher tumour-burden on initial examination (mean volume 884.8 ml) compared to non-type-1-deletion patients (mean volume 357.1ml; p=0.0087) and an increased tumour-growth over time (mean growth/year deletion 61.4 ml vs non-deletion 9.2 ml; p<0.0001). Tumour-burden and tumour-growth were significantly higher for type 1 deletion patients in all age subgroups (all p<0.05).

Conclusion:

NF1 caused by type 1 deletions leads to a high tumour-burden and an increased tumour-growth in affected patients.

Limitations:

The relatively small number of patients studied.

Ethics committee approval

The study was approved by the local ethics review board.

Funding:

No funding was received for this work.