The gold standard evaluation of treatment response in high grade gliomas (HGG) is radiographic assessment of changes in tumor size measured in 2 dimensional structural T1 and T2/FLAIR MR images. Both the McDonald criteria from 1990 and the more recent RANO criteria are based solely on structural imaging and are the most widely used in studies today despite severe limitations¹. Even though median overall survival (OS) in glioblastoma (the most frequent HGG) being less than a year, some responders to therapy live significantly longer and 2 year OS in 20-25% of the population has been reported². Advanced functional MRI techniques are increasingly used in the assessment of HGG and most tumor protocols now include either a dynamic contrast enhanced (DCE) or dynamic susceptibility contrast (DSC) series. There is evidence of both DCE and DSC imaging being able to predict OS in HGG patients^3,4. However, the reported studies vary in timing of examinations relative to treatment schedule, statistical methods and metrics uses. Further, most studies are retrospective in nature and few have investigated the difference in DCE vs. DSC in ability to predict outcome.
To this end we hypothesize that advanced MRI is superior to structural imaging in predicting OS in HGG patients. We further wanted to assess the optimal MRI time-window for early prognosis in HGG and investigate the predictive value of DCE versus DSC in a comprehensive tumor protocol. A prospective study including structural and dynamic images in 27 patients with confirmed HGG (WHO grade III and IV) was performed. Imaging was done before radiochemotherapy (RCT) start (baseline) and every 2 weeks for 8 weeks (five scans) and subsequently every third month. The study protocol included pre- and post-contrast T1, FLAIR, DCE (3D-SR) and DSC (2D SE-EPI). Imaging was performed at 3T (Philips Achieva, Netherlands).
Whole tumor and edema region of interests (ROIs) based on T1 contrast enhancement and FLAIR was produced and approved by an experienced radiologist. Patient specific carry-on individual arterial input functions from the first scan was used for both DCE and DSC, the extended Tofts model was used for the DCE analysis⁵. Whole tumor and edema volume and percentage change from baseline was estimated from the tumor and edema ROIs derived from the structural imaging. Whole tumor median and 90 percentile values for K^trans, v_p, CBF and v_e from the DCE data and normalized and absolute CBV and CBF from the DSC data were analyzed for absolute values and percentage change from baseline (δ).
The subjects were separated in two groups based on short and long OS. Mann Whitney U test was performed with Bonferroni correction for multiple comparisons. Receiver operating characteristics (ROC) curves were investigated for clinically relevant cut-off values for all significant parameters. Log rank test and Kaplan Meyer plots for survival was then produced based on the cut-off values. P<0.05 was set as level of significance and tumors smaller than 0.5 mL were excluded from the analysis.Mean OS was 765 days (min 111 – max 1698) and long OS was defined as survival for more than 750 days (12 patients). Median δCBF and δKtrans from the DCE analysis was significant two weeks after RCT (scan 5). No parameter was found significant at baseline or during RCT (scan 1-4). No DSC parameter was found significant before 6 months after completion of RCT. No structural imaging parameter was found significant. Cut-off values from the ROC curve analysis were conservatively chosen for maximum specificity. An increase in δCBF of more than 15% was associated with a significantly lower OS (p=0.008) (fig2), no significance was found for a single value of δK^trans (p=0,081).

There are three main findings from this study;

1) DCE perfusion MRI is superior to gold-standard structural imaging for early assessment of prognosis in HGG.
2) Optimal timing for prognosis prediction in our material was two weeks after RCT. A baseline scan before start of RCT is needed to assess the change in percent for each parameter (δ) due to the high variation in absolute values between patients.
3) DCE was superior to DSC for early prognosis prediction. No significance for DSC parameters was found before 6 months after RCT. This questions the necessity of both T1 and T2 perfusion in a HGG tumor protocol. The DSC was done using a SE-EPI sequence benefiting from fewer artifacts at the cost of a lower SNR. The use of GRE-EPI could potentially provide higher sensitivity for DSC at the expense of more artifacts from surgical intervention.