Session Item

The difficulty of defining and delineating the clinical target volume for high-grade gliomas (HGGs) is a well-recognized problem within the ESTRO community which lead to the formation of a working group working group aiming at replacing the cumbersome manual process of defining the CTV with a largely computerized process that allows user adjustments. Within this framework, the present work introduces a model for determining the tumor invasion outside the observable gross tumour volume (GTV) and its capacity of predicting the outcome of the treatment and hence guiding the definition and delineation of CTV for HGGs.

A model for the tumor invasion of the normal tissue using a Fisher-Kolmogorov equation which describes tumor cell concentration in space and time by accounting for cell proliferation and cell diffusion was applied on a cohort of 93 HGGs patients. The infiltration of the tumor cells into the normal tissue was quantitatively described by defining volumes encompassed by isocontours with different cell densities assuming that the density of the cells defining the GTV is 8000 cells/mm3, the same as the detectability threshold on MRI scans. The correlation between the volumes defined by different isocontours and the treatment outcome expressed as overall survival (OS) time was sought in terms of ROC curves and Kaplan-Meier survival analysis and compared with the predictions made based on the GTV volume.

Among the volumes defined by isocontours, those corresponding to 1000 cells/mm³ (Fig. 1.a) and 2000 cells/mm³, respectively, resulted in the highest correlation with the OS as rendered by the ROC analysis (Area -Under-the-Curve, AUC, of 0.767, p-value < 0.001) comparable with the predictions based on the GTV (AUC = 0.726, p-value < 0.001). The Kaplan-Meier survival analysis, however, using the 1000 cells/mm³ isocontour (Fig 1.b) and the ROC optimal cut-off volume (i.e. Youden threshold) for patient group selection rendered a p-value < 0.0001 and a hazard ration (HR) of 2.7 while for the GTV (Fig 1.c) the predictions were much poorer from the statistical point of view (HR = 1.6 and p = 0.029).

Figure 1: a) Example of a GTV and isocontour corresponding to a predicted tumor cell density of 1000 cells/mm3. The simulated infiltration appears anisotropic as it conforms to natural barriers and has a preference of spreading via white matter. b)  Kaplan-Meier survival analysis when using a binary survival classification based on volume of GTV, and c) volume of V1000. There is greater separation between survival curves when using the V1000 as parameter, than when using the GTV, indicating that V1000 is a better predictor of overall survival.

The volume defined by the isocontour of 1000 cells/mm3 obtained by modelling the tumour invasion is a stronger predictor of overall survival than the GTV indicating the importance of using mathematical models for cell invasion to assist the definition of the clinical target distribution for HGGs patients.