Dynamic contrast enhanced MRI (DCE MRI) for Phase I anti-angiogenic trial: Comparison of the transfer constant (Ktrans) to blood flow and permeability derived by a distributed parameter model

Abstract

3514 Background: The distributed parameter (DP) model is a DCE-MRI model that enables derivation of blood flow and capillary permeability-surface area product (PS). We aim to compare the DP model to Ktrans in the assessment of angiogenesis assessment in a Phase I anti- angiogenic trial. Methods: Twenty-seven evaluable patients from a completed phase I trial (ABT-869) with 3 dose escalations formed the study population. All patients underwent DCE-MRI at baseline, Day 3 and Day 15 with temporal resolution of 4 seconds. Gadolinium concentrations were estimated using dual flip angle method. Ktrans, normalized IAUC60, permeability-surface area product derived from DP model (DP-PS), and blood flow derived from DP model, were calculated. Patients demonstrating progressive disease in first 2 evaluations scans (cycle 2 or 4) based on RECIST criteria were considered progressors and all other patients, non-progressors. Receiver operating curve (ROC) analysis was performed. Correlation with area under the curve extrapolated to infinity (AUCinf) was analyzed. Results: There is good correlation between DP-PS and drug exposure above threshold (threshold concentration 80 ng/mL; time threshold 3.5 hours) measured by AUCinf (Spearman's coefficient −0.556, p = 0.007). Although slightly weaker correlation is found with normalized IAUC60 (Spearman's coefficient −0.537, p = 0.01), the difference is not significant. (p = 0.9334) There is poor correlation for Ktrans (Spearman's coefficient −0.128, p = 0.570). ROC analysis for predicting progressors versus non-progressors showed a higher ROC area for DP-PS compared to Ktrans (0.869 versus 0.417, p < 0.001) and normalized IAUC60 (0.869 versus 0.643, p = 0.1543, not significant). Using a 29.45% drop from baseline at Day 15 to predict non-progressors, the sensitivity of DP-PS is 64.3% and the specificity is 100% whereas the sensitivity of Ktrans is 28.6% and the specificity is 66.7%. In 7 cases, DP-PS correctly predicted the eventual outcome whereas Ktrans predicted the direct opposite. Conclusion: Permeability-surface area product derived from DP model shows better correlation with drug exposure and may predict patient outcome better than Ktrans. Author Disclosure Employment or Leadership Consultant or Advisory Role Stock Ownership Honoraria Research Expert Testimony Other Remuneration Abbott Laboratories Abbott Laboratories Abbott Laboratories