Do perfusion and diffusion MRI predict glioblastoma relapse sites following chemoradiation?

Fiche publication


Date publication

octobre 2016

Journal

Journal of neuro-oncology

Auteurs

Membres identifiés du Cancéropôle Est :
Pr NOEL Georges, Pr TRUC Gilles


Tous les auteurs :
Khalifa J, Tensaouti F, Lotterie JA, Catalaa I, Chaltiel L, Benouaich-Amiel A, Gomez-Roca C, Noël G, Truc G, Péran P, Berry I, Sunyach MP, Charissoux M, Johnson C, Cohen-Jonathan Moyal E, Laprie A

Résumé

To assess the value of T2* dynamic-susceptibility contrast MRI (DSC-MRI) and diffusion-weighted imaging (DWI) to predict the glioblastoma relapse sites after chemoradiation. From a cohort of 44 patients, primarily treated with radiotherapy (60 Gy) and concomitant temozolomide for glioblastoma, who were included in the reference arm of a prospective clinical trial (NCT01507506), 15 patients relapsed and their imaging data were analyzed. All patients underwent anatomical MRI, DSC-MRI and DWI before radiotherapy and every 2 months thereafter until relapse. Voxels within the sites of relapse were correlated with their perfusion and/or diffusion abnormality (PDA) pretreatment status after rigid co-registration. The relative cerebral blood volume (rCBV) and apparent diffusion coefficient (ADC) were used as biomarkers. Several PDA areas were thresholded: hyperperfused voxels using a 1.75 fixed rCBV threshold (HPt); hypoperfused (hPg) and hyperperfused (HPg) voxels using a histogram-based Gaussian method; diffusion-restricted voxels (DRg); and HPg voxels with diffusion restriction (HPg&DRg). Two sets of voxels (2,459,483 and 2,073,880) were analyzed according to these thresholding methods. Positive predictive values (PPV) of PDA voxels were low (between 9.5 and 31.9 %). The best PPV was obtained with HPg&DRg voxels within the FLAIR hyperintensity, as 18.3 % of voxels without initial PDA were within relapse sites, versus 31.9 % with initial PDA (p < 0.0001). This prospective study suggests that DSC and/or DWI-MRI do not predict the glioblastoma relapse sites. However, further investigations with new methodological approaches are needed to better understand the role of these modalities in the prediction of glioblastoma relapse sites.

Référence

J. Neurooncol.. 2016 Oct;130(1):181-192