A simulation study of methods for handling disease progression in dose-finding clinical trials

TL;DR

This paper evaluates methods for managing disease progression within the DLT observation window in oncology dose-finding trials, using simulations to compare their effectiveness in improving trial efficiency.

Contribution

It introduces and compares new practical methods for handling disease progression in the TITE-CRM framework, addressing a key challenge in oncology dose-finding studies.

Findings

Methods differ in defining evaluable patients and handling incomplete data.

Simulation results show varying performance depending on disease progression scenarios.

Proposed methods improve trial efficiency and accuracy in dose estimation.

Abstract

In traditional dose-finding studies, dose-limiting toxicity (DLT) is determined within a fixed time observation window where DLT is often defined as a binary outcome. In the setting of oncology dose-finding trials, often patients in advanced stage of diseases are enrolled. Therefore, disease progression may occur within the DLT observation window leading to treatment discontinuation and rendering the patient unevaluable for DLT assessment. As a result, additional patients have to be enrolled, increasing the sample size. We propose and compare several practical methods for handling disease progression which occurs within the DLT observation window, in the context of the time-to-event continual reassessment method (TITE-CRM) which allows using partial observations. The methods differ on the way they define an evaluable patient and in the way incomplete observations are included. The methods are illustrated and contrasted in the context of a single simulated trial, and compared via simulations under various scenarios of dose-progression relationship, in the setting of advanced soft-tissue sarcoma.