STEEP: Statistically Efficient Methods for Precision Medicine Trials

Background: All anti-cancer treatments must go through clinical trials for thorough testing before they can be made available widely for treating patients.

Using biomarkers (e.g., genetic aberrations), patients can now be stratified into small subgroups that may receive different benefits from the same treatment. A novel approach called ‘basket trials’ can evaluate a new treatment in several patient subgroups at once. Such patient subgroups typically share the same genetic make-up of the tumour, which is targeted by the new treatment.

Statistical design and analysis for basket trials feature ‘borrowing of information’ across patient subgroups.

However, having multiple patient subgroups to evaluate the efficacy of one treatment may see positive effects claimed more often when they are in fact not clinically meaningful.

In addition, advanced statistical methods that enable adaptation (e.g., allocating patients to a better-suited treatment) are needed for both ethical considerations and accelerating the process of drug development.

Aims: This fellowship aims to develop, as well as to promote the implementation of, advanced statistical methods for precision oncology trials that support decisions to be made in an adaptive manner.

Specifically, this project will propose statistically efficient designs that can 1) control the risk of incorrectly claiming efficacy when there is none, 2) allocate more cancer patients to receive a better-suited treatment, 3) supplement the basket trial data (especially for rare cancers) with historical controls or real-world evidence, 4) enable adding new patient subgroups or new control treatment(s) to an ongoing basket trial.

Methods: My group will work closely with UK clinical trials units (CTUs) that run precision oncology trials to motivate the methodological research for flexible basket trials. Bayesian methods will be developed for analysing trial data at various interim analyses and the trial completion.

Collaborate with patient representatives will also be pursued to incorporate consideration of patient benefits into the decision making.

By doing three placements at The Institute of Cancer Research and several visits to the Cardiff/Southampton CTUs, we aim to identify potential obstacles for the practitioners to use adaptive methods. Easy-to-implement methods and softwares will be developed accordingly to meet their needs. How the results of this research will be used: Statistical methods will be published in leading journals.

Open-source software and training material will be developed to improve the uptake of advanced methods for implementation in the UK CTUs and worldwide. This will ultimately allow more efficient clinical evaluation and better treatments for cancer patients.