Coming to terms with randomisation

Hugh Laurie as House (Source: Wikipedia)

In my previous post I wrote a bit about RCTs and how they help us to know what does and doesn’t work. This is a very powerful tool for finding out what works but it requires a lot of effort in paperwork, analysis, recruiting participants and controls and in monitoring them. What’s nice is that this effort isn’t always needed. There are times when you can do RCTs with almost no effort and still learn a great deal. The problem is that there can be reluctance to do so.

The inspiration from this is drawn from chapter 5 of Bad Pharma, so if you want more details then you should go read them there. One of the projects Ben Goldacre is working on is taking data from UK medical records and interrogating it to generate new information. This data is already recorded routinely so it presents a great opportunity for a RCT. With all the framwork in place there doesn’t have to be any special effort to conduct research, just some simple randomisation. The problem is that to do that currently means complying with a multitude of regulations, delays and consent forms that destroy the effectiveness of the system.

These regulations are important for most trials as they look at new therapies and so consent and oversight for patient’s well-being is essential. However there are multiple drugs available for a specific treatment and at times we don’t know which one is more effective. In those situations a doctor can prescribe either one with no questions asked. To know for sure which one is better we need to do a RCT. As all the patient information is recorded anyway, we just need to randomise the drug that is given. But that’s where the problem lies. Giving a patient either drug is fine but randomising which drug the patient gets is not fine, even though we have no idea which one is better.

If that seems a bit ridiculous, it’s because it is. We don’t want a random diagnosis but at times we just don’t know enough for our decision to be better than random chance. We need to accept that that is the case and come to terms with random chance being a part of our lives. I first became aware of this strange situation of choosing vs randomly assigning drugs from the show House. The situation there was not exactly the same and was fictional but stories are a way we can try understand the world.

In season 1 episode 4 House has to deal with an infection that is killing two babies. They put the babies on two antibiotics because they aren’t sure what’s making them sick and that covers all their guesses. Then the babies’ kidneys start failing due to side-effects of the drugs. Both drugs damage the kidneys but they still don’t know which one will fight the infection. House decides to take one baby off one drug and the other off the other drug. That way they can find out which one is helping them and reduce the strain on their kidneys.

At this point we wind up with the aforementioned paradox (or a similar enough situation). House doesn’t know which drug is helpful. If he keeps using both the babies will die. If he stops the wrong one the babies will die. He is allowed to give them either drug. But, when he wants to to give them separate treatments for the same condition he is called in by the hospital administrator to explain himself. There she and a lawyer tell him it’s not acceptable to decide who lives and dies by the flip of a coin and that the parents must consent and know the babies are getting different treatment. Despite saying he can’t experiment on babies, the administrator eventually relents and let’s him do it.

The objections always struck me as odd. If he doesn’t know what is best to use then it shouldn’t matter which drug he does use. The overall chance of survival is the same. If you split the drugs then one dies and one lives. If you take off one drug you have a 50/50 chance that both die or both live. Two surviving plus two dead divided by two possible outcomes is still an average survival of one. There isn’t a logical objection, and it’s the same when randomising treatments when we really don’t know which is best.

As it turns out both House’s guesses were wrong so neither drug was helping but they only found out after one of the babies had died. But if they were both on one drug and one died they would still have switched to the other drug and the second baby would also have died while they waited to see if it would work. By saving time and testing both drugs simultaneously they were able to save more babies. There were other less seriously infected babies who were also at risk and whose fate depended on House figuring out what was infecting them. In real medicine those extra babies are all the future cases where knowing which drug is best can be the difference between life and death.

At times we just have to come to terms with random chance. When we can’t make an informed choice then randomly assigning medicine can actually be the best option. If you don’t find me convincing then you can listen to Ben Goldacre explain his project, RCTs and randomising drugs in the video below.


1 thought on “Coming to terms with randomisation

  1. Pingback: Two years, still going strong | Evidence & Reason

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s