Tumor volumes measurement (3D) versus Response Evaluation Criteria in Solid Tumors (RECIST version 1.1, 1D) and Choi criteria (C) in assessing response of gastrointestinal stromal tumors (GIST) to imatinib (IM).

Abstract

10059 Background: Assessing antitumor effect of treatment is particularly important to guide treatment changes and to assess novel treatments in early clinical trials. We compared the sensitivity of 3D and 1D measurements in detecting size changes of liver metastases from GIST and assessing response to treatment. Methods: IM was given until progressive disease (PD) as assessed by RECIST. We evaluated liver metastases from 58 GIST patients (pts) by manual (2 independent readers) and semi-automated measurements on CT-scans at baseline, after 6 and 12 months (mo) of IM. Bland-Altman plots were used to evaluate agreement between readers and techniques. Percentages of cases with size changes ≥20% assessed by 3D and 1D were compared by McNemar's test (Pietanza et al, ASCO 2008). The ability of 3D, 1D and C to evaluate response was based on 2 target lesions. Minimally relevant changes by RECIST (+20/-30%) were related to a volume scale to obtain corresponding change cut-offs of +73/-65% and +20/-30%, considering lesions as sphere- or ellipsoid-like, respectively. Association with overall survival (OS) was determined by log-rank test. Results: Mean inter-observer variability for manual measurements was very low: 0.77% for 3D and 0.06% for 1D. 3D detected a size change ≥20% more frequently than RECIST at 6 mo (82 vs 44%, P=0.0001) and 12 mo (29 vs 12%, P=0.0005). Manual and semi-automated methods provided clinically consistent similar measures (variability 0.55 and 0.28% for 3D and 1D respectively, with R2 = 1.0 and 0.99, respectively). At 6 mo, 3D was more sensitive than RECIST in assessing partial response (PR) (60 vs 14% pts, P<0.0001) by using ellipsoid cut-off. OS correlated with PD at 6 mo (P=0.02 for both 3D and 1D and 0.01 for C) and 12 mo (P=0.001 for 3D, 0.002 for 1D, not significant for C). Conclusions: In detecting PR, 3D (ellipsoid cut-off) and C are superior to RECIST 1.1. 3D is a better discriminator of size changes than RECIST 1.1 in GIST, rendering it an interesting method to apply in early clinical trials.