New adaptive platform designs for clinical trials in an emerging disease epidemic

Lead Research Organisation: University of Warwick
Department Name: Warwick Medical School


Before they are made available for general use in the population, drugs are evaluated in clinical trials to determine that they are safe and effective. The planning of these clinical trials to ensure they are able to provide definitive answers usually takes many months, with decisions being carefully depending on the drugs to be tested, the number of patients to be included, the trial duration required to recruit this number of patients and the data collected to assess whether or not the drugs work as hoped.

When a new disease, such as COVID-19, emerges, there is a both a desire to start clinical trials as soon as possible and considerable uncertainty over exactly how these trials should proceed. The number of people likely to be infected, the best way to measure treatment effectiveness in the new disease, and even the best treatments to test, may all be unknown.

A useful approach in such a setting is an adaptive design. This allows a clinical trial, once started, to be modified in a number of ways. One type of adaptive design is a platform trial design, in which additional drugs can be included after the trial has commenced, possibly at the same time as drugs previously under investigation are dropped from the trial if the data suggest that they are not sufficiently promising, giving considerable flexibility. Although such designs are not new, a number of statistical questions remain over the best approach to ensure that the risk of erroneously indicating that a new drug is effective, is kept acceptably low. Additionally, previously proposed methods do not usually provide the level of flexibility desired for clinical trials in an emerging disease.

An adaptive trial can be stopped if the conclusions of the trial are sufficiently clear. In order to control the type I error rate, however, most trial designs require the maximum number of patients that can be included in the trial to be specified in advance. In most disease settings, this presents no challenge as there is good information on the number of patients likely to be recruited. This is in contrast to the setting of an emerging disease, when there can be considerable uncertainty regarding the extent and duration of an epidemic. In this case it might be desirable to continue to recruit as many patients as possible while an outbreaks persists. Similar uncertainty can exist regarding the number of experimental treatments that might be available to be tested in a platform trial, with type I error rate control using existing methods again requiring this to be specified in advance.

Standard methods for clinical trial design and analysis require specification in advance of a primary endpoint used for the evaluation of experimental treatments. In an emerging disease, there may be uncertainty regarding the best endpoint, and it might be desirable to plan and start a clinical trial based on one endpoint, but to modify this as the trial progresses based on data from the trial as well as information from external sources.

This project will develop novel statistical methods to solve the three challenges of uncertainty over the number of patients, the number and timing of treatments to be evaluated, and the best endpoint to be used in the evaluation. This will provide valid analysis methods for clinical trials that have the flexibility needed to enable clinical investigators to adapt the trials in reaction to new knowledge in a developing epidemic setting.

To ensure that the methods we develop are widely disseminated and have maximum impact on clinical trial practice, we will organise a workshop with key stakeholders including clinicians with expertise in emerging infectious diseases, clinical trialists with experience in this area, statisticians with expertise in adaptive trial designs, and relevant regulatory body representatives. Finally, we will produce a webinar to for patients and the general public to explain platform designs for clinical trials.

Technical Summary

Adaptive designs include interim analyses, at which a clinical trial may be stopped or modified, without inflating the type I error rate. One adaptive design of recent interest is a platform trial, which assesses multiple experimental treatments with the ability to add further treatments into the study or drop those that are ineffective. Such designs have been recommended by WHO and NIHR for evaluation COVID-19 treatments, and are being used in ongoing trials, allowing newly-proposed therapies to be tested as quickly as possible.

Although adaptive platform trial designs offer valuable flexibility, given the considerable uncertainty regarding disease progression and patient numbers in an epidemic of a new disease, it is desirable to make clinical trials even more flexible. This is shown by ongoing COVID-19 trials in which the maximum number of patients to be included and duration of the trial, the number of experimental treatments and when these will be available, and the primary endpoint that will be used to evaluate treatment effectiveness, have all been left unspecified. With this flexibility some ongoing trials can only approximately control type I error rates.

This project will develop new statistical methodology to allow the desired flexibility whilst maintaining the scientific integrity of the trial by ensuring control of the type I error rate. Specifically, we will develop methods that allow the maximum sample size and the number and timing of treatments to be added to the trial to be unspecified and allow the trial primary endpoint to be changed.

Research findings will be disseminated via a workshop where we will obtain input from a wide range of methodological and clinical colleagues to facilitate implementation. Consensus will sought where possible and areas of remaining controversy identified. We will also produce a webinar to explain and raise awareness of platform trials to stakeholders including patients and the general public.


10 25 50