RandomiseMe: our fun new website that lets anyone design and run a randomised controlled trial.

December 16th, 2013 by Ben Goldacre in bad science | 1 Comment »

Screenshot 2013-12-16 14.46.15Catching up and blogging this year’s activities: here’s a fun website I made with my friend Carl Reynolds, fellow doctor behind NHS HackDays (where nerds who love the NHS build useful tools). RandomiseMe lets you design and run randomised controlled trials, either on yourself, or on your friends. You can do a trial to see if your new trainers let you run faster than your old ones, find out if cheese gives you nightmares, or club together with friends and work out which kind of gloopy abdomen baste is best at preventing stretch marks in pregnancy. 

I think it’s useful, in the sense that you can use it to design RCTs that help answer questions that are relevant to your life. It’s also a good way of thinking through the strengths and weaknesses of trial design. For example, it would be nice to do a blinded trial of different running shoes, but that would either be very expensive, or involve falling over a lot. Does the absence of blinding matter? It depends on the trial. If you want to do a trial on different strategies to prevent stretch marks, what outcome are you going to measure, how will you standardise it, will you ask people to post pictures? And so on.

Thinking out loud, I think it opens up the possibility to do RCTs that reflect the public’s interest, but wouldn’t get funding. Crucially – tho you get into all kinds of tricky territory with medical trials – there is also the possibility to crowdsource RCTs on medical questions that would be impractical through conventional channels. For example: people inevitably use pain relief with their young children, but there is currently some uncertainty about which of paracetamol and ibuprofen is best to use, for avoiding asthma in later childhood. Given the fact that asthma is pretty uncommon (1 in 100 kids, or 10 in 100 kids, depending on how and who you ask) you’d need to recruit tens of thousands of parents into a trial to answer this question. They’d all need to believe you, when you say that there isn’t currently a clear answer; and they’d all need to care enough to participate in your RCT, and commit to buying either calpol or ibuprofen in the supermarket for the next five years, then reply to your email asking if their kid has asthma. I think you’d struggle to recruit participants like that on a maternity unit, but the venn diagram overlap – recognise uncertainty, inquisitive enough to try resolving it, have a newborn kid – is perhaps easier when you’d crowdsourcing participants online.

Anyway: I think the site is halfway towards being good, but it definitely needs some tweaking, so it would be great if you could use it and tell us how you get on, and specifically what we can do to make the workflow clearer. We’ve had some fun times testing it out with kids designing RCTs as part of classroom activities, and there are some researchers aiming to modify the system and use it to run RCTs recruiting participants in the real world, if you’d like to discuss doing more of this, get in touch.

Lastly, if you’re interested in how the site came about: I wrote this article in the Guardian pointing out that it’s easy to run RCTs on yourself and your friends, NESTA said “go on then here’s some money”, Carl joined my non-profit company Better Data to organise things, and we designed and built it with his nerdy coder friends. Better Data is hatching some more good stuff that will be coming soon.


Screenshot 2013-12-16 14.46.15




If you like what I do, and you want me to do more, you can: buy my books Bad Science and Bad Pharma, give them to your friends, put them on your reading list, employ me to do a talk, or tweet this article to your friends. Thanks! ++++++++++++++++++++++++++++++++++++++++++

One Response

  1. nekomatic said,

    December 18, 2013 at 3:25 pm

    That looks like a lot of fun! The help section is a bit sketchy at the moment, if you don’t mind me saying, and maybe not pitched at the right level – e.g. the bit on ‘what is a power calculation’ launches straight into null and alternative hypotheses. I assume you’d welcome crowdsourced contributions to improve this?