Layman’s abstract: Developing a Bayesian adaptive design for a Phase I clinical trial: a case study for a novel HIV treatment

Every Friday on Statistics Views, we publish layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.

The Open Access article featured today is from Statistics in Medicine: ‘Developing a Bayesian adaptive design for a phase I clinical trial: a case study for a novel HIV treatment’ by Alexina J. Mason, Juan Gonzalez-Maffe, Killian Quinn, Nicki Doyle, Ken Legg, Peter Norsworthy, Roy Trevelion, Alan Winston and Deborah Ashby.

Read the layman’s abstract below.

Clinical trials play a crucial role in the development of new medicines. There are several types of clinical trials. When a medicine is first tested in humans (men and women), this is referred to as a Phase I study. After this Phase II studies and then Phase III and IV studies are conducted. Progress through these phases is dependent upon success at the previous phase.

Phase I studies test a drug for its safety and its efficacy (or how the drug performs). Safety, when using the drug in humans, is crucial. Adverse events (or side effects), especially serious side effects, may lead to a clinical trial being halted. In addition, Phase I studies must provide information about the correct dose of a drug, including the minimum and maximum doses. This guards against 1) side effects which may be caused by too high a dose and 2) the drug not working because the dose is too low.

Phase I studies ask a lot of crucial questions and therefore the quality of their design is critical to getting the right answers. Poor design increases the chances of 1) stopping new research into a promising new drug and 2) wasting resource on further research into a drug that does not work. It also has implications for the success of any follow-up study, because the dosages used in the next phases of development are chosen using information about the doses investigated in Phase I.

Standard designs for a Phase I trial use simple rules. The study participants are grouped, and each group is given a pre-specified dosage, generally in ascending order. For example, the rule may start by allocating the lowest dose to the first group of participants and then move up the scale. However, it will only ascend if there are no safety issues with the lower dosages.

An alternative and more flexible approach uses a design which adapts to the initial results of the study. These adaptive designs use information already collected during the trial to help choose better performing dose levels for the next participant. At present, adaptive designs are mostly used in cancer drugs trials, but there are potential benefits in many other settings.

This paper discusses an adaptive design for a Phase I trial for a novel HIV treatment. Such trials are important because new drugs are needed to treat an ageing population with HIV and other medical conditions. The likely performance of a range of adaptive designs is compared to a standard design for seven different test scenarios using simulation. The results show that there is no optimal design for all circumstances. After much consideration of the benefits and risks of choosing an adaptive design over a more standard design, the authors agreed it was likely that an adaptive design would provide the best information for making decisions about follow-up studies.

Developing an adaptive design is time-consuming. It should be carried out in parallel with other aspects of the trial setup. To be successful all members of the protocol development team must be fully involved, as input is needed from HIV clinicians, statisticians and experts in clinical trial design. Importantly, there was also involvement from patient groups. This effort is justified if a more efficient design is chosen. For future trials, this paper shows how to develop an adaptive design for HIV treatments, and is of important relevance in other clinical areas.

The full article is available as Open Access:

Developing a Bayesian adaptive design for a Phase I clinical trial: a case study for a novel HIV treatment

Alexina J. Mason, Juan Gonzalez-Maffe, Killian Quinn, Nicki Doyle, Ken Legg, Peter Norsworthy, Roy Trevelion, Alan Winston and Deborah Ashby.

Statistics in Science, Volume 36, Issue 5, 27 November 2016, pages 754-771, https://doi.org/10.1002/sim.7169