The Number Needed to Treat

I've written in the past about using data in understanding medical recommendations, particularly in the context of cancer screening, what with new recommendations about PSA (and mammogram) testing. You can see my earlier posts here, here, and here.

Today I am adding a new (for non-statisticians) concept, the number needed to treat. Put simply, the Number Needed to Treat is the number of people who need to take a medication, or have a procedure, in order, statistically, for someone to have a good outcome. Here's a great video, from the website TheNNT.com, explaining the concept:

Here's a link to the NNT's explanation in writing, for those of you who prefer to learn by reading.

Decisions about what procedure to have, or whether to start taking a medication, require the exercise of judgment. Adding the number needed to treat gives you one more piece of information to consider. If the number needed to treat is large, and the side effects sound daunting, then perhaps you might make a different judgment than when the number needed to treat is smaller. Among many other great features of the site is a quiz. The quiz gives you basic information about the medication (how the medicine is delivered, its cost) and the numbers (how many trials, how many patients, how many trials showed a benefit). Unlike other quizzes, this one is designed to help you practice weighing different factors about a medication. Because value judgments are involved, your answers may differ from the site's answers, but there is always an explanation.

There's a lot more to this site that I will take up in future posts. For now, this concept is definitely worth a look.

Cancer screening, questioned

Here's a thoughtful article by Gina Kolata from Saturday's NY Times, assessing the spate of new guidelines suggesting that less cancer screening may be a better public health approach. I've discussed this issue in earlier posts, particularly my review of "Overdiagnosed," here (and the article quotes the book's principal author Dr. H. Gilbert Welch). Cost and a new understanding of cancer--including the fact that many cancers do not grow at all or grow slowly--have led to the recommendations of less screening.

Medical statistics and decision-making

Back in April, I discussed Welch, Schwartz, and Woloshin's excellent book "Overdiagnosed: Making People Sick in the Pursuit of Health" in a post. The authors assume readers will want to focus on the numbers, and provide a clear guide to interpreting medical advice that comes your way. I was reminded of it several times in the past week. First the US Preventive Services Task Force issued a draft statement recommending against PSA screening for  asymptomatic men.

Second, the NY Times Magazine discussed PSA testing in an article on Sunday titled, "Can Cancer Ever Be Ignored?" FWIW, the article quotes Welch as saying, “The European trial says 50 men have to be treated for a cancer that was never going to bother them to reduce one death. Fifty men. That’s huge. To me, prostate screening feels like an incredibly bad deal.”

Third, last week I read Jerome Groopman and Pamela Hartzband's newest book "Your Medical Mind: How to Decide What is Right for You." That book is not a numbers book by any means, but it does provide several spectra intended to help the general public figure out how to respond when a doctor proposes tests, medication, or other procedures. Written in the clear style we have come to expect from Dr. Groopman, it is pitched at a level you can take in while distracted by the emotions and time pressures of a medical crisis.