What Were They Thinking? Pharmacologic priors implicit in a choice of 3+3 dose-escalation design

TL;DR

This paper explores how the implicit pharmacologic assumptions in 3+3 dose-escalation trial designs can be inferred, highlighting the underlying beliefs about safety and pharmacology that influence trial structure.

Contribution

It introduces a method to reverse-engineer the implicit pharmacologic priors embedded in a common oncology dose-escalation design.

Findings

Implicit pharmacologic priors can be inferred from trial design

Design choices reflect underlying safety and pharmacology assumptions

Method provides insights into trial safety expectations

Abstract

If explicit, formal consideration of clinical pharmacology at all informs the design and conduct of modern oncology dose-finding trials, the designs themselves hardly attest to this. Yet in conducting a trial, investigators affirm that they hold reasonable expectations of participant safety - expectations that necessarily depend on beliefs about how certain pharmacologic parameters are distributed in the study population. Thus, these beliefs are implicit in a trial's presumed conformance to a community standard of safety, and may therefore to some extent be reverse-engineered from trial designs. For one popular form of dose-escalation trial design, I demonstrate here how this may be done.