What are randomised controlled trials and why do they provide such good evidence?

I’m going to introduce you to the key features of the randomised controlled trial and explain why such trials provide the best evidence to inform treatment decisions for the millions of people around the world living with disease. Suppose we have a new treatment and we want to find out whether or not it’s effective, so whether or not it makes people with a specific disease or condition feel better. One thing we could do is to treat a handful of patients with the treatment and see what happens, recording whether or not they get better over time. But there are a few problems with this. Some people may get better over time anyway because of the natural course of the disease or condition.

So, for example, with the common cold, the majority of people will feel better after five days, whether or not they’ve had any treatment. Sometimes people perceive themselves to feel better just because they know they have received treatment, so they have an expectation that there will be an improvement even if the treatment itself has no medical effects. This phenomenon is what we call the placebo effect. The natural course of a disease or condition and the potential of a placebo treatment effect mean that if we simply look at what happens to a small group of patients receiving a new treatment, we may think that the treatment is effective when in actual fact, it isn’t.

So the first thing we need to improve our study is to have a comparison or a control group, a group of patients who do not receive the new treatment that we’re trying to evaluate. This control group should receive the current standard treatment for that disease or condition so that the new treatment can be compared directly with what we already know works. If there is no standard treatment available for this group of patients, then we might use a sham or placebo treatment. This is a dummy treatment that has no medical effects but could still induce a placebo effect. After treatment has been received, we assess some measure of how patients are outcome of interest.

For example, if we were interested in a treatment effect on pain, we would measure the level of pain in each patient after treatment and then compare pain scores of patients receiving the new treatment with those receiving the control treatment. Having a control group means that we can take account of the fact that some patients may naturally get better over time or feel better simply because they are being treated. So now, if we observe that the treatment group is doing better than the control group, at the end of the study, we can be more confident that the treatment is beneficial to patients. So how do we decide who gets the new treatment and who gets the control?

Well, we could leave it up to the health professionals and patients themselves to decide, but both parties are likely to have strong preferences for which treatment they prefer. And this can lead to big differences in the characteristics of patients who receive the new treatment compared to those receiving the control. So, for example, a doctor may prefer for those with more severe illness or who are older to receive the standard treatment which we know more about and for the healthier or younger patients to receive the new treatment. But this is not a fair comparison. We’re not comparing like with like the younger, healthier patients are likely to have better health outcomes, irrespective of whether the new treatment is better or not.

So if there are more healthy young patients in the new treatment group than the control group, we’re unlikely to get the right answer. We’re likely to conclude that the new treatment is better than it really is. And our estimate of how effective the new treatment is compared to the control will be inaccurate or biased. In addition to things like age and severity of disease, there are also likely to be many other factors that we don’t even realise affect outcome. But that may differ between the patients receiving each treatment. So the best way to achieve a fairer comparison across treatments is to randomise patients into the study to receive either the new treatment or the control treatment.

So each patient recruited to the study has the same chance of receiving either treatment as the next patient, normally a 50 50 chance of receiving each treatment. So, for example, for each patient, we could throw a coin heads. They get the new treatment, tails, they get the control treatment. Nowadays, the process of randomisation is done by a computer programme as this avoids any human error or manipulation and enables a record to be kept out of the process. If randomisation is done properly, patients allocated to each treatment group will be similar in terms of all known and unknown factors that affect the outcome. We have a fair comparison. The groups are similar in terms of everything other than the treatment itself.

If we now observe that the treatment group is doing better than the control group at the end of the study, we can be even more confident that the treatment is beneficial to patients and that we have a true unbiased estimate of treatment effect. And this is a randomised controlled trial, the inclusion of a control group and the randomisation of patients to the treatment groups is what makes it the best study design for determining whether or not a new treatment works or if it is as good as or better than the current standard treatment.

Finally, whilst I’ve illustrated the importance of randomised controlled trials within the context of patients and health care, they are similarly used to reliably answer crucial questions in social care and public health.