SO-36 An immune-related gene expression profile predicts the efficacy of adding atezolizumab to first-line FOLFOXIRI/bevacizumab in metastatic colorectal cancer: A translational analysis of the phase II randomized AtezoTRIBE study

The AtezoTRIBE study demonstrated that the addition of atezolizumab to first-line FOLFOXIRI/bevacizumab prolongs PFS of mCRC patients, but a clinically modest benefit was observed among patients with proficient mismatch repair (pMMR) tumours. Therefore, identifying a pMMR subgroup able to achieve benefit from immune-checkpoint inhibitors (ICIs) is a crucial challenge of translational research. We investigated the predictive role of an immune-related gene expression signature (IO score), which reflects the presence of infiltrating inflammatory cells versus a differentiated stromal microenvironment. AtezoTRIBE was a phase II comparative trial in which mCRC patients, unselected for MMR status, were randomized 1:2 to receive first-line FOLFOXIRI/bevacizumab (control arm) or FOLFOXIRI/bevacizumab/atezolizumab (experimental arm). RNA was obtained from FFPE blocks of pre-treatment tumour specimens from 142 (65%) out of 218 enrolled patients. RT-qPCR was performed using DetermaIO™, to assess mRNA expression of a 27-gene targeted panel. In each sample the IO score was calculated according to the established pan-cancer IO algorithm. The predefined IO cut-point (0.09) was applied to dichotomize tumours as IO + or IO -. To identify patients deriving maximal benefit from the experimental arm, an exploratory analysis was conducted to find an optimized IO score cut-point (IO OPT), by using a method based on maximizing the log-rank statistics for comparing the two groups. The identified IO OPT cut-point was then adopted to dichotomize tumours as IO OPT+ or IO OPT-. IO score was successfully determined in 122 (86%) cases. Thirty-three (27%) tumours were IO +. No differences between IO + and IO - tumours were observed in terms of baseline clinical and molecular features and clinical outcome (mPFS: 14.4 vs 13.6 mos; HR:0.84 [95%CI:0.53-1.33], p=0.468). An interaction between IO status and treatment effect was reported (p for interaction=0.066), with higher PFS benefit from the experimental arm among patients with IO + (HR:0.39 [95%CI:0.15-1.02]) than IO - tumours (HR:0.83 [95%CI:0.50-1.35]). In the pMMR tumours (n=110), a similar trend was observed (IO + tumours [n=30]: HR for PFS 0.47 [95%CI:0.18-1.25]; IO - tumours [n=80]: HR for PFS 0.93 [95%CI:0.56-1.55]) (p for interaction=0.139). In the overall population (n=122), the computed IO OPT cut-point was 0.277. Sixteen (13%) tumours were IO OPT+. When compared with IO OPT- ones, IO OPT+ tumours were more frequently TMB-high (p=0.007), had longer PFS (mPFS: 14.8 vs 13.3 mos; HR:0.50 [95%CI:0.28-0.87], p=0.053) and derived significantly higher PFS benefit from the experimental treatment (HR in IO OPT+ 0.10 [95%CI:0.02-0.52] vs HR in IO OPT- 0.85 [95%CI:0.54-1.33], p for interaction=0.004). Similar results were observed in the pMMR subgroup (n=110), where the computed IO OPT cut-point was 0.304 and twelve tumours (11%) were classified as IO OPT+. The investigated IO score signature using the predefined cut-point may help to predict benefit from the addition of atezolizumab to first-line FOLFOXIRI/bevacizumab in mCRC, even within pMMR tumours. The exploratory IO OPT cut-points should further be validated in independent mCRC cohorts.