Early change in tumor size from waterfall plot analysis and RECIST response as predictor of overall survival (OS) in advanced, chemotherapy-refractory colorectal cancer (ACRC): NCIC CTG/AGITG CO.17 study.

Abstract

3602 Background: The CO.17 trial demonstrated that the addition of cetuximab monotherapy (C) to best supportive care (BSC) improves OS and progression-free survival (PFS) in patients (pts) with ACRC. We compared tumour measurement as a continuous variable and visualized by waterfall plot vs. categorized as response based on RECIST as potential early predictors of OS benefit. Methods: Of the 572 randomized, 216 pts on C and 142 on BSC had tumor assessment at baseline and 8 weeks and are included in this analysis. Response at 8 weeks was assessed using standard RECIST criteria and categorised as partial response, stable disease or progressive disease. Continuous tumour size measurement was also assessed using waterfall plot, The absolute difference in logarithm transformed sum of the longest tumor diameters (LSLD) between 8 weeks and baseline were calculated for the entire group and the KRAS wild-type (WT) and mutant (MUT) subgroups and correlated with OS in a multivariate analysis including difference in LSLD, RECIST response and other prognostic factors as covariates. Results: The increase of tumor size (cm) from baseline of pts treated by C was significantly smaller than those that received BSC (mean difference in LSLD 0.02 vs. 0.23; p < 0.0001), and this remained significant after adjusting for baseline prognostic factors (p<0.0001). In a multivariate analysis of pts treated by C, difference in LSLD was significantly associated with OS (HR 14.84, per ln(cm) of increase, 95% CI 4.31 to 51.08; p<0.0001), whilst RECIST defined response was not. Results were similar for WT pts: HR 14.30 per ln(cm) of increase in LSLD, 95% CI 1.32 to 96.5; p=0.03. For MUT pts, neither change in LSLD nor RECIST response was associated with OS. Conclusions: Change of tumor size at 8 weeks following commencement of C using a waterfall plot analysis was a better predictor of OS than standard RECIST categories. This may be a better early signal of treatment efficacy for molecular targeted therapies in ACRC.