Skip to 0 minutes and 13 seconds IAN ROBERTS: If you want to know the effects of a treatment on health outcomes, ideally what you want to do is you want to get two groups of people who are otherwise the same, apart from the fact that one got the treatment and one didn’t. Now, getting two groups of people who are otherwise the same, is not so straightforward. And the best way we’ve got of doing that is by randomly allocating people into the group that gets the treatment, and the group that doesn’t. And if chance and chance alone determines who goes into which group, then on average those two groups will be the same, apart from the treatment.
Skip to 1 minute and 3 seconds And so randomisation is really special because it gives you two groups of patients that are identical, on average, apart from the treatment. And so any differences in health outcomes can be attributed to the treatment.
Skip to 1 minute and 18 seconds JUDITH GLYNN: And if you could just explain the use of a placebo.
Skip to 1 minute and 21 seconds IAN ROBERTS: A placebo helps to ensure validity in evaluation by making sure that the two groups of patients are treated exactly the same way, an outcome is measured in exactly the same way. And it reduces bias, and it helps you to get really reliable results. Now importantly, it doesn’t mean to say that you are denying the group that get the placebo good quality care. So everybody can get all the treatments that we know are effective, and on top of that, one group gets the treatment being tested, and the other group gets the placebo. So giving a placebo doesn’t imply that you’re not caring for patients.
Skip to 2 minutes and 12 seconds I think especially in the Ebola context, people have got confused about this and think it’s unethical to give people placebo. Well, it’s not unethical to give placebo because you’re not denying them anything effective. You’re just denying them treatment that you’re uncertain about. And there shouldn’t be anything wrong with that. If we don’t know if a treatment works, then you’re not missing out on anything.
Skip to 2 minutes and 37 seconds JUDITH GLYNN: So specifically for Ebola, why are these randomised controlled trials needed?
Skip to 2 minutes and 44 seconds IAN ROBERTS: Because there are some things that we know are effective, but there are lots of uncertainties. We haven’t had a lot of experience in caring for patients with Ebola. And so the way we normally improve the quality of care for patients is by an iterative process of doing randomised controlled trials. Right, this treatment works. Now, let’s give that to everybody. This could improve the outcome further. Well, let’s do a randomised trial of that, and then you find if that works or not. But many of the things that we do to patients are not effective, or they can be ineffective. And many treatments can be harmful.
Skip to 3 minutes and 25 seconds So randomised controlled trials are really essential for separating the good treatments from the useless, or harmful treatments. Everybody should get known effective treatments. But if we’re uncertain about the effectiveness of a treatment, then it’s the right thing to do scientifically. It’s the right thing to do ethically. It’s the right thing to do all around. And in fact, it’s routine that patients are enrolled in randomised controlled trials in high income countries. The way we improve health care in high income countries is by doing randomised controlled trials. We really need to improve the care of Ebola patients, and so we need to do randomised controlled trials. Why should West Africans be denied the very important benefits of good quality clinical research?
Skip to 4 minutes and 15 seconds Why should they have to put up with poor quality research? The sorts of research designs that are being proposed, like before and after studies, which often get the answer wrong. And if you get the answer wrong, it’s a catastrophe because you can treat lots of patients with treatments that do more harm than good. You can really damage people. You can cause a lot of suffering. It’s not impossible to do randomised controlled trials in Ebola.
Skip to 4 minutes and 46 seconds If you can give them a treatment, then you can give half of them the treatment. If you can give them an untested treatment to all of them, you can give an untested treatment to half of them. And giving it to half of them, enables you to find out whether it works or not. So I’m very sceptical about the arguments that people say that it’s not practical.
Skip to 5 minutes and 7 seconds JUDITH GLYNN: What about the issues of getting informed consent when you’ve got someone who’s coming in, who’s very sick, and there’s no relatives with them?
Skip to 5 minutes and 13 seconds IAN ROBERTS: OK. So patients who are really ill are the exception to the general rule of informed consent. So you don’t always need to get informed consent to put patients in a randomised controlled trial. Now, a lot of people don’t realise that, but it’s a principle that’s been widely accepted all around the world. It’s acknowledged in the Declaration of Helsinki that if we want to find better treatments for patients who are critically ill, in acute life threatening situations, we have to allow entry into randomised controlled trials without informed consent. Now, we do that in many of our trials because our patients often have acute severe head injuries. So they’re bleeding into their brain.
Skip to 6 minutes and 4 seconds They can’t give informed consent themselves because they’re unconscious. It’s not appropriate to wait for a relative to come in because all that time they’re bleeding into their brain. So ethics committees look at the protocol, and they consider the situation. And they accept that a waiver of informed consent, in those situations, is appropriate. Otherwise, we’ll have no effective treatments for patients in that situation.
Skip to 6 minutes and 30 seconds JUDITH GLYNN: If your daughter had Ebola, would you be happy to enrol her in a randomised controlled trial? Wouldn’t you want whatever experimental treatments were out there, whatever was looking the best at that time? Or would you accept a 50% chance that she might get a placebo?
Skip to 6 minutes and 45 seconds IAN ROBERTS: I would want her to get all of the treatments that have been shown to be effective. Now, that’s first and foremost what I would want. And the sad thing is that in Ebola patients, they’re not getting all of the things that we know are effective. So the reason that patients in high income countries have a low case fatality rate, and people in low income countries, in West Africa, have a high case fatality rate, is that they’re not getting the basic managements of fluids and electrolytes that you would get in an intensive care unit, or just in a hospital in the west. So they’re not getting known effective treatment.
Skip to 7 minutes and 27 seconds So for my daughter, I would want her to get a known effective treatment. And then, would I allow her to be randomised to a treatment that we didn’t know was effective or not? Well, absolutely yes. This idea of compassionate care that you throw untested drugs at vulnerable patients, and you can call that compassionate. It’s really ridiculous.
Skip to 7 minutes and 50 seconds JUDITH GLYNN: The standard way of assessing whether a new drug works is to do a randomised controlled trial, usually including a placebo in the arm that doesn’t get the new drug. Are there situations in which randomised controlled trials don’t work, or you can’t do them, it’s not appropriate?
Skip to 8 minutes and 4 seconds PETER SMITH: I think there are situations where they’re probably not appropriate. They’re rare. If you’ve got a therapy that you’ve really got enormously strong prior evidence that this is going to be a miracle therapy from animal studies, or other preclinical studies, then you might want to try it in a few patients just to see if it does have that miracle effect. And if it does have, as it were, an all or nothing effect of patients who usually die, survive, then the evidence from that sort of study may be sufficient to kind get a therapy into use. The classic example, that’s always given, is that nobody’s ever done a randomised controlled trial of whether parachutes work.
Skip to 8 minutes and 51 seconds Most therapies are not like that, we’re normally looking for a smaller increase in survival, or the cure rate of particular diseases. And then randomised controlled trials are absolutely relevant. The problem comes, I think, and we’ve hit this in the Ebola epidemic, when you’ve got a disease which has a very high fatality rate. And this is true for Ebola. It’s not true for some cancers, in sort of late stage cancers. If you want to do a trial of a new therapy there, is it going to be possible to go to patients and say, we’ve got this experimental therapy? It may work, it may not work.
Skip to 9 minutes and 34 seconds And we’d like to do a trial, but we’re going to randomise you to receive a placebo, or this experimental therapy. Would you be prepared to enter that trial? I think many patients, many of us, I think, if we had a disease which had a 70, or 80, 90% fatality rate, we’d probably want to take the chance with an experimental therapy. Rather than standing a 50 or 30% chance of receiving a placebo, which we can be fairly confident is going to have no effect. So I think it’s perfectly ethical to try to do trials in those circumstances. But practically, I think it may be very difficult to actually persuade patients that the trial is the way forward.
Skip to 10 minutes and 19 seconds JUDITH GLYNN: The whole purpose of a placebo controlled trial is that you have a very fair comparison. You know that the patients are the same apart from the drug or the placebo. If you’re not using that design, how do you make the comparison?
Skip to 10 minutes and 34 seconds PETER SMITH: It becomes much more problematic. And essentially, it depends a bit on availability. For some of the therapies that are being used for Ebola, they aren’t in plentiful supply. So not all the patients can receive them. And so some patients will receive them, some won’t, and you can compare the survival of those two groups. However, with some of the other sort of antiviral treatments, they are in fairly plentiful supply so everybody can be treated. Whether or not that does any good remains to be seen. But the comparison group then is more problematic.
Skip to 11 minutes and 11 seconds And essentially what is being proposed is a historical control, where you take patients who were treated before this therapy became available, and you compare their survival rate with patients who receive the therapy. There are problems with interpreting those sort of data, but it may be all that can be done to actually evaluate these therapies.
Skip to 11 minutes and 34 seconds JUDITH GLYNN: Knowing what the case fatality rate is like without these new treatments is actually quite difficult. It’s varied from place to place, there’s problems to know who’s included. Since the underlying case fatality rate without these new treatments might be anywhere between about 40% and 70%, how are we really going to know if a new treatment works unless it’s wonderful?
Skip to 11 minutes and 54 seconds PETER SMITH: Well, in order to use a new treatment you have to set up procedures in a treatment centres so that you can actually collect data in a way in which you can get a good estimate of what the survival rate would be. And you’d want to do these trials in treatment centres where these procedures have really already been set up in terms of normal care. So you have got a good estimate of what the survival rate of patients going to a particular centre is.
Skip to 12 minutes and 21 seconds Of course, it’s always possible that at the same time you put in an antiviral treatment, say, there’ll be other changes that were made with respect to treatment procedures that will make it very difficult to interpret that difference. But that’s, unfortunately, a difficulty that I think we’re going to have to cope with, because I just don’t think, in many circumstances, it’s going to be practical to do the classic randomised controlled trial, which has an arm in it where patients don’t receive the experimental therapy. If the effect of the antiviral is small, then it may be very difficult to distinguish that effect from bias.
Skip to 13 minutes and 0 seconds If the effect is large, then one might be more confident that this is a true effect of the intervention.
Skip to 13 minutes and 8 seconds JUDITH GLYNN: You mentioned for some new treatments there might be quite small quantities available. In that situation, doesn’t it become more ethically possible to randomise because some people are not going to get them anyway?
Skip to 13 minutes and 19 seconds PETER SMITH: Absolutely. And I think in the circumstances where you’ve got, as it were, two patients who were otherwise similar, but you’ve only got therapy for one, the most ethical way to actually allocate that therapy is to, as it were, toss a coin to randomise. Whether that’s going to be possible in the circumstances of the treatment centres, where to some extent the staff working in the treatment centres are concerned primarily with giving the best possible care to their patients. And imposing a research component on that care, at least some view that as potentially disruptive with respect to normal care. I think, obviously, if one sets up a research study, it is potentially disruptive with respect to normal care.
Skip to 14 minutes and 13 seconds So there have to be very careful procedures put in place to ensure that there isn’t that sort of disruption. So it may be possible to randomise in some circumstances. I think that it should certainly be tried. But if it’s not possible, it’s still going to be necessary to find some way of evaluating these experimental therapies.
Skip to 14 minutes and 35 seconds JUDITH GLYNN: And are there any innovative designs that have come up to allow them to be evaluated without doing trials in the usual way?
Skip to 14 minutes and 43 seconds PETER SMITH: There are some so-called adaptive designs, where essentially all patients who come to a treatment centre will be treated. And as each patient either survives or dies by either 14 days or 21 days, then they will be plotted, and looking at the survival rate continuously. And if the survival rate– and this could be sort of formally worked out statistically– looks very poor, then treatments that are not doing well can be stopped very quickly. If, on the other hand, patients survive very well, then again, that can be accelerated, and more patients can be put on to that therapy, or in other centres.
Skip to 15 minutes and 27 seconds Or if it goes on and it gets to a stage where it’s not very bad, but it’s not very good. And you might get to a stage where you decide, well, we really don’t know. And that might be the time that you could more strongly argue for a more conventional randomised controlled trial.
Should randomised controlled trials be used to evaluate new treatments?
As potential new therapies for Ebola become available the question of how best to test them has been hotly debated. Should they be subject to randomised controlled trials in the usual way, or are there compassionate or practical grounds on which they should simply be used, relying on historical comparisons to assess the effects?
In this video, Professor Ian Roberts and Professor Peter Smith put forward opposing arguments. After watching the video and reading the text in this step, go to the next step to consider a scenario and cast your vote depending on the decision you make.
A randomised, placebo controlled trial for a treatment is an experiment in which one group of patients, the treatment group, gets an experimental treatment, and the other, the control group, gets a placebo. The use of a placebo means that neither the patients nor those treating them know who is getting the active treatment. Patients are allocated to the treatment or control group using a random process.
Randomised controlled trials (RCTs) are recognised as the least biased way of testing new treatments because they ensure that the patients in the two groups are the same, apart from the new treatment. It is important to note that both groups receive the same supportive care available. Even though RCTs are the standard approach used to test new treatments, there are arguments against using them in this particular situation for both ethical and practical reasons. These are summarised by a group of authors (including Professors Peter Piot and David Heymann) in a letter to the Lancet.1
The arguments in the letter are:
With conventional care, mortality is so high that ‘it is problematic to insist on randomising patients to it when the intervention arm holds out at least the possibility of benefit.’
Western medical workers were not randomised before receiving experimental therapies.
There is a precedent for not using a control group in studies of cancers with poor prognosis.
In a situation where there is a lack of trust in health care ‘insisting on RCTs could even worsen the epidemic, by undermining trust in the Ebola treatment centres that are central to containing it.’
Conducting RCTs may not be practical in the fraught conditions of the epidemic.
Other methods might give results more quickly.
© London School of Hygiene & Tropical Medicine