Tumor assessments according to RECIST versus volumetry: Impact on early changes in tumor size in metastatic colorectal cancer (mCRC).

Abstract

e14043 Background: We previously proposed an algorithm for reconstructing the true volume of a tumor lesion by using the longest diameter (LD) and the longest orthogonal diameter (LOD). The following analyses aimed to replicate the previous findings, and evaluate the agreement between RECIST-based tumor assessments (TA) and volumetry in the quantification of early changes in tumor size. Methods: mCRC patients (pts) from the FIRE-3 trial (FOLFIRI + cetuximab vs FOLFIRI + bevacizumab) underwent routine staging using contrast-enhanced CT scans. At least 2 hepatic target lesions per pt were defined for 2 raters at baseline and measured every 8 weeks using semi-automated volumetry (Siemens Syngo Via Oncology, Siemens Healthcare) and manual measurement of the LD according to RECIST 1.1, and the LOD. Agreement between the tumor volume algorithm and volumetry for each rater and the corresponding inter rater agreement (IA) were investigated using Bland-Altman (BA) plots with limits of agreement (LOA). Agreement between the RECIST-based TAs and volumetry was quantified by intraclass correlation (ICC). Both raters and the statisticians were blinded to the treatment given. Results: 222 target lesions from 25 pts were measured at 99 TAs. BA plots indicate that the volume algorithm showed negligible constant bias for both raters (0.09 with p=0.25 and 3 observations outside LOA, and -0.05 with p=0.41 and 6 observations outside LOA). IA with regard to the volumetry showed a smaller constant bias (-0.16 with p=0.17, no observation outside LOA) compared with the volume algorithm (-0.31 with p=0.04, one observation outside LOA). The relative change in tumor size between baseline and first TA was investigated. The ICCs between RECIST-based TAs and volumetry for the 2 raters were 0.79 and 0.73. Data in relation to treatment efficacy will be presented. Conclusions: The algorithm for estimating the tumor volume was validated as shown by the negligible bias and low number of measurements outside the LOA in the BA plots. The moderate ICCs for early changes in tumor size indicate differences between RECIST- and volume-based TAs and suggest that volume measurements discern early changes in tumor size more precisely.