VOLUMETRIC TUMOR MEASUREMENTS ARE SUPERIOR TO 2D BIDIRECTIONAL MEASUREMENTS IN THE EVALUATION OF IDH INHIBITION IN DIFFUSE GLIOMAS: EVIDENCE FROM A PROSPECTIVE PHASE I TRIAL OF IVOSIDENIB

Abstract Since IDH mutant (mIDH) low-grade gliomas (LGGs) progress slowly and patients have a relatively long survival, testing of new therapies in clinical trials based solely on survival can take more than 20 years. Guidance on therapeutic evaluation using LGG RANO criteria recommends serial bidirectional (2D) measurements on a single slice; however, questions remain as to the best approach for evaluating LGGs in clinical trials including use of volumetric (3D) measurements, which would theoretically allow for more accurate measurements of irregular shaped lesions and allow readers to better assess areas of change within these tumors. A total of 21 (out of 24) non-enhancing, recurrent mIDH LGGs with imaging pre- and post-treatment enrolled in a phase I, multicenter, open-label study to assess the safety and tolerability of oral ivosidenib (NCT02073994) were included in this exploratory ad hoc analysis. 2D bidirectional and 3D volumetric measurements were centrally evaluated by one of 3 radiologists at an imaging CRO using a paired read and forced adjudication paradigm. The effects of 2D vs. 3D measurements on progression-free survival (PFS), growth rate measurement variability, and reader concordance and adjudication rates were then quantified. 3D volumetric measurements had significantly longer estimates of PFS (P=0.0181), more stable (P=0.0063) and considerably lower measures of tumor growth rate (P=0.0037), the highest inter-reader agreement (weighted Kappa=0.7057), and significantly lower reader discordance rates (P=0.0002) with comparable recommended LGG RANO 2D approaches. In summary, 3D volumetric measurements are better for determining response assessment in LGGs due to longer PFS and more stable measures of tumor growth rates (i.e. less “yo-yo-ing” of measurements over time causing fewer erroneous calls of progression and more accurate growth rates), highest inter-reader agreement, and lowest reader discordance rates. Future studies will focus on validating this in a larger cohort and determining whether these measurements better reflect clinical benefit.