Progression-free survival in intention to treat populations versus total KRAS populations in patients treated for metastatic colorectal cancer: A pooled review

Abstract

4054 Background: In treatment trials of patients with metastatic colorectal cancer (mCRC), KRAS mutation status of tumor samples has been retrospectively demonstrated to be predictive of treatment benefit. Historically, clinical trials have not required tissue samples to be tested for KRAS mutation status as a condition of enrollment. Therefore, KRAS analyses have been based on available tissue samples representing only a portion of patients retrospectively analyzed for KRAS status and correlated with treatment end- points. Methods: A weighted analysis of pooled data was performed using six recently presented or published clinical trials of targeted therapy in mCRC which tested for the association of KRAS status with Progression Free Survival (PFS). The goal of the analysis was to determine whether there is a significant difference in PFS between the intention to treat (ITT) population and KRAS population in both the treatment and control arms. Results: The total ITT population of the pooled studies is 3864, and for the total KRAS population, 2295; a 59.4% retrieval rate (range 28–92%) for tissue samples available for KRAS analysis. The weighted Δ PFS across all arms between the ITT population and the KRAS population was 0.2 months with a range of 0–0.7 months. Of the 12 subgroups (6 control and 6 treatment arms), five had no difference in PFS between ITT and KRAS evaluable populations at all, and two additional subgroups demonstrated a difference PFS of only 0.1 months. The two studies with the lowest tissue retrieval rates (28% and 45%) had the largest Δ PFS. Conclusions: There is no meaningful difference in the PFS between the ITT and KRAS populations based on our analysis of pooled data. The difference in PFS was greatest in the two studies with the lowest rate of retrieval of tissue samples for KRAS testing. As such, subgroup analysis is better able to estimate and reflect the ITT population if a higher percentage of samples is able to be obtained. Further, our results suggest that there is not an inherent systemic bias influencing any potentially observed differences in PFS. Tissue samples should be required for all patients entering a clinical trial to avoid this issue and to make retrospective analysis more valid. [Table: see text]