Adaptive designs for phase I trials 1

Published on September 29, 2016   31 min

Other Talks in the Series: Adaptive Clinical Trial Design

Other Talks in the Series: The Risk of Bias in Randomized Clinical Trials

Hi, my name is Anastasia Ivanova from the Department of Biostatistics, University of North Carolina at Chapel Hill. And today, we're gonna talk about Adaptive Designs for Phase I Trials. These designs were developed and are used in Phase I trials in oncology. Some of these designs can be used in Phase I and mostly Phase II trials in areas other than oncology. Since these designs are mostly used in oncology, I will focus on oncology applications throughout this lecture.
In oncology, we usually consider several fixed doses from three to eight. And we assume that the probability of toxicity is non-decreasing with the dose. Here, on this plot, I have a curve showing the probability of toxicity as a function of dose. The goal of a Phase I trial in oncology is usually to find the Maximum Tolerated Dose or MTD. The MTD is defined as a dose with a certain probability of toxicity, usually 20%. In oncology, the primary outcome in dose-finding studies is actually the Dose Limiting Toxicity, but I will use these two terms interchangeably DLT or toxicity.
So as I said, the Maximum Tolerated Dose is defined as the dose where the probability of toxicity is equal to a certain number but you know this number as Γ. And Γ is usually 20% in Phase I trials in oncology.
Why do we need to find the MTD in oncology trials? Usually, we assume the therapeutic effect of a drug is increasing with the dose. Therefore, lower doses are usually not efficacious. Due to toxicity, we cannot administer higher doses to most of the subjects. Therefore, the goal is to find the dose with a tolerable toxicity rate. This dose then is investigated in a subsequent Phase II trial. In non life-threatening diseases, the MTD is the highest dose in the range of doses that are further investigated in the Phase II trial.
So how can we design a trial to find the MTD? How can we allocate subjects to estimate the MTD and to find the MTD among the several doses we consider? One possible design is the parallel assignment of subjects with an equal number of subjects assigned to each dose. For example, if we have 6 doses, we can assign 6 subjects to each dose for the total of 36 subjects.
This design will not be suitable for oncology trials because it's not ethical to assign subjects to very high doses with possibly high-toxicity probabilities. In oncology, we start with lower doses and gradually escalate to higher doses. One possible solution is the escalation design, where say, we can start with dose 1, assign 6 patients to that dose, then observe the toxicity outcomes. If we only observed 0, 1 toxicities, then we can assign the next cohort of 6 subjects to the next higher dose. If we observe 2 or more toxicities, then we stop the escalation and declare the dose right below the current one as our estimated MTD. If we continue in this manner, the estimated MTD will approximately have a toxicity probability of about 20%.
One might argue that assigning six subjects to the same dose is also not ethical because we're exposing six subjects at once to a potentially dose with a high toxicity probability. A solution is to assign three subjects first, and then depending on the outcome, potentially assign three more. And this design is called the Traditional or Standard or 3+3 design. This design is often used in the oncology studies. In fact, about 99% of oncology Phase I trials use the Traditional design.

Adaptive designs for phase I trials 1

Embed in course/own notes