Save a Life: Write about the "number needed to harm"
Everyone could benefit from taking cholesterol-lowering statins. Who wants gunk clogging up their arteries, right? And even if your cholesterol is already low, you may gain some wonderful side benefit.
That is the overwhelming message driven home by a combination of marketing campaigns and overly enthusiastic health reporting.
Last month, researchers poked a hole in that balloon with a study in the British Medical Journal. As Kristina Fiore at MedPage Today wrote in a wonderfully well-documented story:
Claims of unintended benefits of statins appear to be largely unsubstantiated and known risks -- including liver and renal problems and myopathy -- confirmed, British researchers concluded based on findings from a large prospective open cohort study. The study involving more than two million people, "largely confirmed other studies that reported no clear association between statins and risk of cancers," Julia Hippisley-Cox, MD, and Carol Coupland, MD, of the University of Nottingham in England reported online in BMJ.
One of the great things Fiore did was talk about the number needed to harm (NNH.) Antidote explained the concept of number needed to treat in May. Number needed to harm is the reverse. You can find the NNH perfectly explained in the Merck Manual:
To gain a rapid picture of the benefit (B) or harm of a given treatment, it is sometimes useful to calculate the number of people that would need to be treated to save or lose one life. That number is simply the reciprocal of the difference in absolute (not relative) risk (R). Considering again the example of thrombolytic therapy for an acute MI, the absolute risk difference is 12% – 9%, or 3% (B), so the number needed to treat (NNT) is 1/0.03 or 33. In the setting of an acute MI, one would need to treat 33 patients with thrombolytic therapy to save one life.
On the other hand, the absolute risk difference for an intracranial hemorrhage is 1% (R), so the number needed to harm (NNH) is 1/0.01 or 100. One would need to give 100 patients thrombolytic therapy to cause one fatal intracranial hemorrhage. These simple calculations do not weigh the relative value of the benefit and the harm, but the NNT and the NNH can provide a useful snapshot of benefits and risks.
Fiore used both NNT and NNH in her story:
They found that for women at high risk of heart disease, the number needed to treat to prevent one case over five years was 37. For men, it was 33. With regard to esophageal cancer, the number needed to treat to prevent one cancer case was substantially higher -- at 1,266 among women and 1,082 among men. For adverse outcomes among women, the number needed to harm for an additional case of acute renal failure over five years was 434, 259 for myopathy, 136 for liver dysfunction, and 33 for cataract. Those numbers were similar among men, except for myopathy, which was significantly lower at 91.
Readers at low risk for heart disease now have a crystal clear picture of whether they should take statins on the off chance that they might benefit in some other way. You might prevent esophageal cancer, but it's doubtful. In fact, you are much more likely to suffer renal failure, muscle problems (myopathy), liver dysfunction or cataracts.
Duncan Echelson at the Medical Skeptic blog made a good point about my NNT post, saying, "The problem is simple, the vast majority of those with access to NNT statistics have little interest in generating or sharing them. In fact, in many instances, it is to their financial and professional interests to keep these figures either unknown or safely tucked away in publications."
That's all the more reason to make sure you gather and understand the absolute risk numbers so that you can crunch the NNH yourself. And, if you get stuck, call someone in the field who can walk you through the math.
Don't let the numbers get the best of you. Or your audience.
Related posts:
Q&A with John Carey: Peeling back the label on cholesterol drugs
10 Things I Learned at AHCJ 2010
Should Vioxx still be on the market?
Danger! Writing about relative risks can lead your readers astray
At the next pharma love fest, spend some time with absolute risk
Talking Risk: Vioxx, statistics and other complexities
NNT can be TNT for blowing up pharma marketing claims