How should we test potential treatments for Ebola?

A pressing problem and a very modern dilemma, how should we test potential treatments for Ebola? We know that a double-blind randomised clinical trial (RCT) is the gold standard for generating evidence for, or against, a new drug. In an RCT patients are randomly assigned to either the new treatment or the “control” – a treatment that it is to be compared against, often a placebo or “the current standard of care”. By assigning patients randomly we hope to avoid any bias, intentional or not, in which patients get which treatment. Because if, for instance, less-sick patients were preferentially given the new treatment or the new treatment was predominantly tested on patients at a better equipped hospital than those on the control, the patients on the new treatment might have a better rate of recovery that owed nothing to the new treatment.

Where possible these trials are also double blind so that neither the patient nor the doctor looking after them knows whether the patient has been given the new treatment or the “control”. Patients have an extraordinary capacity to “do better” if they, or their doctor, think they are getting a better treatment; this is of course the “placebo effect”.

Scientist in Laboratory

But what about Ebola? Patients are dying every day, can’t we just start giving new medicines and see if the death rate falls? The regulators are saying an RCT is necessary for new Ebola treatments. How can this be ethical? Well, in Forbes (, Mathew Herper reports Ed Cox, director of the FDA’s Office of Microbial Products, making the very important point that as the response to the crisis in West Africa improves, the death rate is already starting to come down ( If a new treatment is simply tested on its own, how will we know that any improvement in the death rate is not simply due to continued improvements in basic care? Do we want to pay what will probably be huge sums of money to a biotech for a medicine that delivers no benefit and may have side-effects of its own? If it were your son or daughter who needed treatment, what would you want?

My preference would be to try to be creative in how these new medicines are trialled. Firstly – start the clinical trial now – before supplies of the new medicines are available, building up a dataset of how many patients survive in the clinical trial setting when given the standard of care (which is likely to be at the upper end of the standard of care across the region and actually be a benefit in itself) so fewer need to be randomised to the control arm as sufficient supplies of the new medicines become available for them to be tested in the trial. A degree of blinding (though admittedly less than perfect) could be kept by keeping secret exactly when new treatments are introduced into the trial, and using a randomisation process from the outset even though to start with there was only one treatment: placebo plus standard of care. Secondly, run this as a single trial to test all Ebola treatments; this will minimise the number of patients who have to be given just the standard of care, allow future new treatments to enter a clinical trial more quickly and more easily compare new treatments against each other and not just the standard of care. Thirdly, this may allow the best use of each treatment to be identified; they may perform differently depending on the strain of the disease, or whether the patient is male or female, old or young, or how long they have had the disease. Where a difference is sufficiently clear it should even be possible to preferentially randomise patients so they are more likely to receive the more appropriate treatment for them. This is not an original nor entirely fanciful idea, the I-SPY 2 breast cancer trial is already being run this way.

There’s probably only one thing we have to give up to be able to take this path – and that’s p-values and analytical control of type-1 error. A trial of this complexity will only be analysable using Bayesian statistics and we will have to rely instead on estimating type-1 error by simulation. And, whisper who dares, I for one think that would be a small price to pay to be able treat the patients in the trial better and develop effective medicines more quickly.

Tom Parke

Tom Parke

Tom Parke has been working at Tessella for over ten years. For a large part of that time he has ...

© Copyright 2017 Tessella
All rights reserved