Targeted immune therapy, such as anti-PD1, has improved response rates and survival outcomes in various oncology indications. Tumor microenvironment biomarkers such as PD-L1 via IHC, the inflammation gene expression profile (GEP) and tumor mutation burden (TMB) are in use or under investigation as enrichment strategies for anti-PD1 therapy. Early clinical trials in the advanced, single-arm setting are sometimes used to determine biomarker cut-points to be evaluated in later randomized studies. Often such cut-offs are selected at the time of an interim analysis of the single-arm study via an evaluation of the biomarker's tumor response clinical utility profile (i.e., quantities such as prevalence, sensitivity, PPV and NPV). Interim analyses, however, suffer from the limitation of incomplete follow-up, where not all responses have yet been observed, and thus, have the potential to bias the selected biomarker cut-point. Here we leverage a competing risks perspective to frame an approach using the observed data at the time of interim analysis to impute patients with unobserved response status, aiming to mitigate bias in the estimation of the response clinical utility profile.