Please, PLEASE use absolute risk data

Posted By


Based on my reading of a New York Times story in my local Star Tribune today, I was ready to blast the Times.

I was reading the story out of the American College of Cardiology meeting on the new data suggesting that statins can reduce the risk of venous thromboembolism.

By how much can statins reduce the risk? Depends on which version of the NYT story you read.

The Times reports:

“With the relatively healthy people in the study, the number of clots was small — 94 total — but the placebo group developed 60 of them, compared with 34 for those taking the statin.”

But the Star Tribune version of the Times story doesn’t include that absolute risk reduction figure and, instead, only lists the more impressive sounding relative risk reduction figure:

“relatively healthy people who took a potent statin were 43 percent less likely than those who took a placebo to get a blood clot known as venous thromboembolism.”

The Wall Street Journal’s simple language was the most helpful for readers:

“34 participants who were taking Crestor developed VTE compared to 60 who were taking a placebo – a small absolute benefit but a relative risk reduction of 43 percent.”

To editors who trim wire stories – or who don’t know to look for absolute data – I can’t state strongly enough how incomplete your stories are if you don’t include absolute risk data.

Borrowing an analogy from Woloshin and Schwartz of Dartmouth, citing relative risk only is like having a coupon for 43 percent off. But you don’t know whether it’s 43 percent off the cost of a Lexus or off the cost of a lollipop.


You might also like


Please note, comments are no longer published through this website. All previously made comments are still archived and available for viewing through select posts.

Marilyn Mann

March 30, 2009 at 7:59 pm

Your point about presenting both relative and absolute risk reduction is well taken. I would like to make a comment about the study results. The researchers classified venous thromboembolism (VTE) events into two categories: provoked and unprovoked. Provoked would be people who had cancer, had just had surgery, or other situations that increase the risk of VTE. Unprovoked would be all the other people. The percentage risk reduction was similar in both groups. That may indicate that people who are at high risk of VTE could benefit from taking a statin. This would be like your 43 percent off a Lexus example. However, I think they need to do a clinical trial in high risk people and see if these results are replicated.

Dan Abshear

March 31, 2009 at 7:49 pm

This statistical gymnastics is more common within members of the pharmaceutical company than many are aware of, I believe.
Other tricks include presenting data to prescribers showing superior results of a drug company’s product over another, with the reference being simply ‘data on file’.
Questionable clinical trials performed for some reason in Italy often are utilized in promotion as well illustrating superiority again over a competitor.