Statistical methods for next-generation clinical trials

Clinical trials are used to test the effectiveness and safety of new treatments. In recent years they have become more and more expensive and have high rates of failure. Methods for reducing the cost of clinical trials and improving the chances of finding the truth are important. New technology means that new approaches to clinical trials are needed. This programme will conduct research into statistical methods for improving clinical trials.

The programme will cover three main areas. The first is research for including the large amount of biological information that can be measured for patients. The purpose of this is that different patients may benefit more from a treatment than others: we would like to be able to find this if it is true. The second is to explore how clinical trials can be more efficient by allowing new treatments to be added as they go along. This will reduce the cost and provide opportunities to ensure patients on a trial get the most suitable treatment for them. A third area is to improve the statistical analysis of trials which use complicated ‘composite endpoints’ which measure several things at once. This reduces the cost of clinical trials.

By working with doctors as well as statisticians we will make sure that this research is used in real clinical trials to benefit patients and decision-makers.

The process of developing and testing a new treatment or intervention takes a long time, a lot of money, and often ends in failure. The high cost is primarily due to expensive clinical trials, which are required to show that the treatment is safe and effective. Improving the efficiency of clinical trials, to make maximum use of limited resources, is a priority research area.
This research programme will develop statistical methodology to ensure that clinical trials have the tools necessary to cope with current and future challenges.
The first area of focus is statistical methodology for best using the increasing amount of high-dimensional biomarker data in clinical trials. With increasing availability of routinely collected biomarker data, modern clinical trials urgently require suitable methods to incorporate them prospectively. Current approaches rely on less efficient statistical methods such as testing each variant one-by-one in a logistic regression, and classifying patients as high-risk if they are positive for a certain number of biomarker. By combining novel adaptive designs and state-of-the-art high-dimensional statistical methods such as Bayesian sparse regression we can improve on this. Through working with clinicians, this novel methodology will be available for use in real trials for areas such as cardiovascular diseases and oncology.
A second area of focus is methodology for ongoing trials that test multiple treatments and biomarkers. In an ongoing trial new treatments and biomarkers are added in continually as treatments are found sufficiently promising to move to phase III trials, or are dropped due to lack of effectiveness. Some real trials are already doing this due to the considerable logistical and administrative advantages. This programme will focus on statistical issues in ongoing trials. This includes how to optimally choose decision criteria for dropping or progressing a drug to phase III and how to optimally plan phase III trials that result. Collaborations with a multidisciplinary set of statisticians and clinicians will allow us to implement developed methods in practice.
A third area of focus is to improve the analysis of trials using composite endpoints. In this case it is possible to gain considerable power by fitting a suitable model to the data. These methods will be applied to a variety of disease areas including oncology trials, rheumatoid arthritis and lupus.