Estimands in oncology early clinical development: Assessing the impact of intercurrent events on the dose-toxicity relationship

The R1 addendum to ICH E9 (E9-R1) provides guidance on the definition of estimands in clinical drug development. While the E9-R1 has seen uptake in randomized late-stage clinical trials, its implementation in early clinical development remains sporadic potentially jeopardizing clarity, consistency, and coherency in early phase. In this article, we call for a more systematic use of the estimand thinking in phase 1 dose escalation oncology trials. In these adaptive trials, the primary clinical objective is usually to characterize the dose-toxicity relationship and to ascertain the maximum tolerated dose (MTD). One estimand of interest is the probability of dose-limiting toxicity (DLT). Intercurrent events (ICE) interfering with the existence or interpretation of DLT outcomes are common in these studies. Three types of ICEs are reviewed in detail: treatment discontinuation for reasons not related to toxicity (e.g. disease progression), treatment discontinuation due to adverse events which would not qualify as DLT, and dose modifications or omissions. The concept of replacement of non-DLT evaluable participants, often used so far, is not an acceptable general solution to ICEs in dose escalation studies. To address clinically relevant questions, adequate ICE handling strategies and estimators aligned to these settings should be used.