Optimal combination of conventional MRI and DCE-MRI parameters to early predict pathologic response to anthracycline-based neoadjuvant chemotherapy for locally advanced high-grade soft-tissue sarcomas - DUMAS - Dépôt Universitaire de Mémoires Après Soutenance Accéder directement au contenu
Mémoires Année : 2017

Optimal combination of conventional MRI and DCE-MRI parameters to early predict pathologic response to anthracycline-based neoadjuvant chemotherapy for locally advanced high-grade soft-tissue sarcomas

IRM conventionnelle et perfusionnelle pour l'évaluation précoce de la réponse des sarcomes des tissus mous de haute grade localement avancés sous chimiothérapie néo adjuvante à base d'anthracycline

Résumé

Purpose : to determine which conventional MRI and Dynamic-contrast-enhanced MRI (DCE-MRI) qualitative, semi-quantitative, quantitative features are significantly associated with histologic response to anthracycline-based neoadjuvant chemotherapy (a-NAC) in locally advanced high grade soft-tissue sarcoma (STS). To estimate inter- and intra-observer agreements of each criterion. Methods: retrospective, monocentric exploratory study. 46 adults with biopsy-proven high-grade STS of trunk walls and extremities, uniformly treated in a French sarcoma reference centre with 5-6 cycles of a-NAC. Radiological evaluation consisted in conventional pre-therapeutic (t0) MRI0 and after 2-3 cycles of a-NAC (t1) MRI1+DCE-MRI1. After 5-6 cycles of a-NAC (t2), 32 patients also had MRI2+DCE-MRI2 for pre-operative evaluation. All had available pathological analysis of surgical specimen with treatment response estimation. Good or poor histologic response (GHR, PHR) was determined depending on % residual viable tumour cells (10% or >10%, respectively). MRI were randomly analysed, two times by two radiologists, blinded to histological results. Conventional MRI lecture assessed tumour maximal size, bi-dimensional diameters, volume, as well as T1 and T2 signals, architecture, margins, oedema, nerve, vessels and bone invasion in a dichotomised fashion. DCE-MRI was analysed on whole tumour surface (ROIs) and on the first area displaying contrast uptake (ROIN): qualitatively (%tumour volume enhanced during arterial phases, curvology), semi-quantitatively (model-free: wash-in (WIn), iAUC90, Peak-enhancement (PE), and their ratio versus healthy muscle) and quantitatively (extended Tofts model: Ktrans and its ratio). Se, Sp, PPV, OR were determined, as well as best cut-offs helped by receiver operating characteristics (ROC) curves analyses. Significant criteria to distinguish GHR and PHR at t1 were used in a logistic binary regression to build a multivariate model to predict response. Inter- and intra-observer agreements were estimated. At t2, correlation between %viable cells and numeric DCE-MRI2 indices were evaluated with uni- and multivariate linear regression. Results: 34 PHR and 12 GHR were included. Gender female was significantly associated with PHR (p=0.015). After 2-3 cycles of a-NAC, none of the change in measurements or on conventional MRI was significantly different between PHR and GHR. PE and WIn-ratio on ROIN were significantly higher in PHR (p= 0.029 and 0.048, respectively). When gender, PE and WIn-ratios variables were considered together, they significantly predicted histologic response (p=0.004), but none individually reached significance. 94.1% PHR were well-classified. 45.5% GHR were ill-classified. AREA under ROC curve of the model was 0.846 IC95%=[0.677-0.994]. At t2: %change of longest dimension, bi-dimensional diameter, volume since t0 were significantly higher in GHR (p=0.012, 0.010, 0.031, respectively). WIn-ratio, PE, iAUC90, Ktrans,, Ktrans-ratio in ROIN (p= 0.012, 0.010, 0.031, 0.022, 0.023, 0.020, 0,009, 0,010, respectively) as well as Ktran, Ktrans-ratio on ROIs were lower in GHR (p=0.043 and 0.021, respectively) . Multivariate linear regression analysis showed that WIn-ratio-ROIn,2 and % change of the longest dimension were significantly correlated with % viable cells on following surgery (r=0.370, p=0.022 and r=0.394, p=0.029, respectively). t2 analysis identified late responders profile in 4/5 ill-classified true GHR. Conclusion: in combination with MR-morphometrics features, DCE-MRI improves response prediction to a-NAC, correlates with % residual viable tumour cells on post-a-NAC surgical specimen, highlights a ‘late-responders’ profile. Estimators of fibrosis and necrosis should be added to the model. Prospective multicentre studies are required for validation.
Objectifs : déterminer les paramètres de l’IRM conventionnelle et perfusionnelle (DCE-MRI, qualitative, semi-quantitative, quantitative) associés à la réponse histologique sous chimiothérapie néoadjuvante à base d’anthracycline (a-NAC) pour sarcomes des tissus mous (STM) de haut-grade localement avancés. Méthode : rétrospective unicentrique. 46 adultes avec diagnostic histologique de STM de haut-grade de la paroi du tronc ou des membres, traités uniformément en centre de référence par 5-6 cycles de a-NAC, avec une IRM conventionnelle pré-thérapeutique MRI0 et après 2-3 cycles MRI1+DCE-MRI1. A 5-6 cycles, 32 ont aussi eu MRI2+DCE-MRI2 pour évaluation préopératoire. Tous ont bénéficié d’une analyse histologique sur pièce opératoire pour déterminer la réponse, bonne ou mauvaise (GHR, PHR) selon le % cellules tumorales viables (seuil à 10%). Trois régions d’intérêt (ROI) tracées : sur l’aire la plus précocement rehaussée de la tumeur (ROIN), sur une surface représentative maximale de la tumeur (ROIS), sur le muscle sain(ROIM). Se, Sp, PPV, OR ont été déterminés ainsi les meilleurs cut-offs. Les paramètres distinguant significativement GHR et PHR à t1 ont été testés dans par régression logistique binaire pour proposer un modèle multivarié de prédiction. Les reproductibilités inter, intra-observateurs ont été calculées. Résultats : 34 PHR, 12 GHR. Significativement plus de femme PHR (p =0.015). A t1 : PE et WIn-ratio dans la ROIN étaient significativement plus élevés chez PHR (p= 0.029 and 0.048). La combinaison genre, PE et WIn-ratio prédit significativement la réponse (p=0.004) mais aucun seul n’atteint la significativité. 94.1% PHR ont été bien classés mais 45.5% GHR mal classés. L’aire de la courbe ROC du modèle était 0.846 IC95 (0.677-0.994), p<0.0001. Conclusion : DCE-MRI améliore la prédiction précoce de la réponse histologique sous a-NAC des STM. Des études prospectives multicentriques sont requises pour validation.
Fichier principal
Vignette du fichier
Med_spe_2017_Crombe.pdf (3.18 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

dumas-01649831 , version 1 (27-11-2017)

Identifiants

  • HAL Id : dumas-01649831 , version 1

Citer

Amandine Crombé. Optimal combination of conventional MRI and DCE-MRI parameters to early predict pathologic response to anthracycline-based neoadjuvant chemotherapy for locally advanced high-grade soft-tissue sarcomas. Life Sciences [q-bio]. 2017. ⟨dumas-01649831⟩
53 Consultations
116 Téléchargements

Partager

Gmail Facebook X LinkedIn More