Number Needed to Treat

As I've attempted to explain in the past, divining the truth and predicting the future is not an easy thing to do, at least when it comes to applying the latest in medical research to your own situation.  When reading the latest headlines from newspapers & journals, we're often buffeted about.  On any given day, we're supposed to do this; the following day, we're not.  A great portion of the blame falls upon our shoulders as, too often, we don't do a good job of explaining observational data vs randomized controlled trials.  The former is only good for developing links, associations & hypotheses but observational data is useless for guidance, no matter how strong or repeatedly we see similar results.  Guidance & proof, at least from a scientific & statistical sense, can only be discerned from randomized controlled trials.

I mention this because over the weekend, I stumbled upon a fascinating website, www.thennt.com, which focuses on randomized controlled trials and calculates the number needed to treat/harm from the available data.  To truly practice evidence-based medicine, we need to be able to explain our recommendations in a fashion that non-statisticians (like myself) can easily understand.  Many of the conclusions at this website run counter to conventional wisdom, mainly due to the number needed to treat, as calculated from the absolute risk reduction, rather than the more impressive relative risk reduction.  There's more information than I can go into in such a brief space, but I would urge you to look over this website at your leisure and reconsider what you're doing to improve your health and that of your patients.