P-262 - Trial-level analysis of early tumor shrinkage and disease control rate as intermediate end points of first-line medical treatment in randomized studies of metastatic colorectal cancer

Abstract

Introduction: Early tumor shrinkage (ETS) is a response-related end point (EP), defined as a radiologic dimensional reduction of at least 20% after 8 weeks of chemotherapy, and it has been shown to be predictive of overall survival (OS) in studies of metastatic colorectal cancer (mCRC). From the available studies and a patient-level meta-analysis it appears that ETS correlates better than overall response rate (ORR) with OS [Sommereir 2014, J Clin Oncol 32:3538]. However, a trial-level pooled analysis concluded that ETS predicted progression-free survival (PFS) but not OS. Disease control rate (DCR) by RECIST appeared as a good intermediate end point of OS from a previous analysis of 11 randomized trials of patients with mCRC who underwent first-line treatment with chemotherapy plus bevacizumab vs. chemotherapy alone, reporting a better performance than PFS [Colloca 2016, Clin Oncol 28:e155-64]. The aim of the present study is to perform a trial level analysis to verify if DCR and ETS after a first-line chemotherapy are related to OS in patients with mCRC receiving various regimens. Methods: After a systematic search of randomized trials reporting ETS, randomized clinical trials (RCTs) were selected whenever they evaluated PFS, ETS, ORR and DCR in relation to OS. Two arms per trial were selected, and the differences in the results of these two arms for every EP (Δ, delta) were calculated. The nonparametric Spearman ρ (r) was used as a measure of correlation between the difference in each end point and the difference in OS. The analysis evaluated the treatment effects on ΔPFS, ΔETS, ΔORR or ΔDCR and on ΔOS, by separate linear regressions for every EP, and the proportions of variability explained (R2trial) on OS for the four EPs (PFS, ETS, ORR and DCR) have been calculated. Results: The systematic review of the literature led to the selection of 12 RCTs (7 phase-3, 5 phase-2). Despite the limitations and possible bias related to the limited number of patients with an ETS report for each study (in total 3117/4327 patients; 72%), ETS have a different performance in the whole sample vs. phase-3 trials (R2trial = 0.172 vs. 0.842), similarly to ORR (R2trial = 0.349 vs. 0.740). Of the response-related EPs DCR is the most closely related to OS (R2trial = 0.541 vs. 0.816). PFS was significantly related to OS when all trials are included but not when phase-3 trials only are considered (R2trial = 0.586 vs. 0.466). Conclusion: In phase-3 trials response-related end points are more accurate in OS prediction and performs better than PFS. However ETS needs a prospective validation, considering its retrospective determination in all available trials. DCR is a very interesting intermediate EP also in studies with chemotherapy plus EGFR-inhibitors or chemotherapy alone and should be explored at pre-established time points.