Abstract Purpose: Tissue derived tumor mutation burden (TMB) of ≥10 mutations/Mb is a histology agnostic biomarker for the immune checkpoint inhibitor (ICI) pembrolizumab. However, the dataset on which this was validated lacked colorectal cancers (CRCs), and there is limited evidence for immunotherapy benefit in CRC using this threshold. Patients and Methods: CO.26 was a randomized phase II study of 180 patients comparing durvalumab and tremelimumab (D+T, n=119 patients) versus best supportive care (BSC, n=61 patients). ctDNA sequencing was available for 168 patients (n=118 D+T, n=50), of which 165 had evaluable plasma TMB (pTMB). Tissue sequencing was available for 108 patients. Optimal thresholds for stratifying patients based on overall survival were determined using a minimal p-value approach. This report includes the final overall survival analysis. Results: Tissue TMB ≥10 mutations/Mb was not predictive of benefit from D+T compared to BSC in microsatellite stable (MSS) metastatic CRC (HR 0.71 [95% CI:0.28-1.80], p=0.47). No tissue TMB threshold could identify a high TMB group that benefited from ICI. In contrast, plasma TMB (pTMB) ≥28 mutations/Mb was predictive of benefit from D+T (HR=0.34 [95%CI:0.13-0.85], p=0.022), as was clonal pTMB ≥10.6 mutations/Mb (HR=0.10 [95%CI:0.014-0.79], p=0.029) and subclonal pTMB ≥25.9/Mb (HR=0.20 [95% CI:0.061-0.69], p=0.010). Higher pTMB was associated with length of time on cytotoxic agents (p=0.021) and prior anti-EGFR exposure (p=2.44x10-06). Conclusion: pTMB derived from either clonal or subclonal mutations may identify a group more likely to benefit from immunotherapy, though validation is required. Tissue TMB provided no predictive utility for immunotherapy in this trial.