A comparison of biologics in first-line advanced colorectal cancer: A Bayesian network meta-analysis of EGFR inhibitors and bevacizumab.

Abstract

543 Background: Adding bevacizumab (B) or EGFR inhibitors (E) to chemotherapy have improved outcomes when compared to chemotherapy (chemo) alone in the first-line treatment of mCRC, however it is unclear which of these combinations is optimal. As the 2 RCTs presented to date were not powered to detect overall survival (OS) benefits and have shown conflicting OS results, a meta-analysis may be beneficial. Methods: We conducted a systematic review of RCTs comparing (1) E + chemo vs. B + chemo (2) E + chemo vs. chemo only, or (3) B + chemo vs. chemo only, using MEDLINE, Embase, Cochrane Central, and ASCO abstracts up to June 2013 with Cochrane methodology. Data on PFS and OS were extracted using the Parmar method. For RCTs involving E, only the K-ras WT data was included. The patient characteristics and outcomes of the reference arms of the RCTs were examined to assess for heterogeneity. Bayesian pairwise and network meta-analyses (NMA) were conducted to estimate the direct, indirect and combined PFS and OS hazard ratios comparing E to B using WinBUGs. Results: Seventeen RCTs (8,048 patients) were identified; 15 of them contained extractable data for quantitative analysis. Direct pairwise meta-analyses (2 RCTs) comparing E vs. B showed that PFS HR=1.00 (95% credible regions (CR): 0.86-1.17) and OS HR=0.76 (95% CR: 0.63-0.92) in favour of E. Indirect comparisons of E vs. B (through the intermediate of chemo only: 5 RCTs comparing E + chemo vs. chemo only, 8 RCTs comparing B + chemo vs. chemo only) showed that PFS HR=1.31 (95% CR: 0.98-1.85) and OS HR=1.06 (95% CR: 0.93-1.22). Combining direct and indirect comparisons with NMA (15 RCTs) showed that the PFS HR=1.10 (95% CR: 1.00-1.21) (trend in favour of B) and OS HR=0.95 (95% CR: 0.85-1.06). Conclusions: The results of direct pairwise meta-analysis, dominated mostly by FIRE-3, suggested E improves OS without PFS benefits when compared to B. However, the results from indirect or combined NMA synthesizing all relevant data from the existing literature did not confirm those findings. The findings of FIRE-3 may be due to chance or trial specific reasons. The results of the upcoming CALGB 80405 will provide further direct evidence to help refine these estimates.