■ site map
BLOG: February 2007
Are the health studies results misleading?In his Time Magazine health article (February 2007), Michael D. Lemonik argues that the usual manner in which results of health studies are presented doesn't convey the whole truth. It may be even misleading.
The reason is that a figure - most often given in percentage points - of how much a certain factor (treatment, nutrient, medication, toxin, etc.) increases the chances of a certain outcome, doesn't say nothing about
the actual probability
To illustrate this, the author cites a 30% heart attack rate reduction with statin drugs - according to a study - as opposed to no treatment. It implies quite significant benefit, but what is needed to judge its actual importance is the basic risk level for heart attack.
The article states that it is so called "number needed to treat" (NNT) which reveals this vital piece of information. It tells how many patients, on average, need to be treated in order to average a single specific effect - in this case benefit - compared to the control group (no treatment, or some other treatment).
For statins, the NNT is, according to the article, about 50 for heart attack. This number tells you that, statistically, statin treatment will benefit one out of every 50 patients. In other words, if your condition is comparable to those of subjects in the study, it would statistically decrease your chances of suffering heart attack by 1/50, or 2%.
Now, there is a huge apparent difference between 30% (heart attack rate reduction) and 2% (your personal statistical risk reduction). While both numbers are correct, the latter is the one that gives you the actual picture. The numbers implicate that your chances of suffering heart attack are about 4.7% with statin treatment, and either 30% or 2% more - about 6.7% - depending on whether you take relative or nominal increase, without any treatment.
This results from a simple formula for the "number needed to treat",
with %CG being the percentage of a specific outcome in the control group (no treatment, placebo, or established therapy) and %EG being the percentage of that same outcome in the experimental group.
But the NNT value alone still hides the important aspect of
what is your risk of being affected in the first place.
It is merely an efficacy indicator for the statistical patient population. NNT would be the same (50) for any 2% nominal difference in the EG/CG efficacy, say, for 1% and 3%, or 50% and 52%. In the first case, the relatively low, but not negligible risk is reduced threefold - definitely worth attention - and in the second one the very high risk is only slightly - statistically negligible - reduced.
Seems that KISS ("keep it simple silly") approach is still the best. Why not to omit still potentially confusing NNT terminology, and
state it simply as it is:
for a given group of patients, statins reduced the chances of heart attack from 6.7% to 4.7%, or from 1 in 15 to 1 in 21. It shows clearly both, degree of benefit and treatment importance (measured by the chances of your health being negatively affected if not subjected to a treatment). Is that too plain and simple for those who conduct medical studies?
Here are the NNTs for some common ailments/treatments you may find interesting (source: Centre for evidence-based medicine).
You've heard so many times doctors (and layman-like) recommending daily aspirin for heart protection. It has become part of the culture. But when you look at the numbers, statistically, it will help only one person in 200. On the other hand, vitamin E supplementation will help 1 person in 7, while exercise will help two out of every five persons.
Why is vitamin E, despite being
30 times more efficient heart protection than aspirin
so infamous? Probably the main reason is that nutrients in general, and their significance in preserving and regaining health are still being mainly ignored by today's "official", business-minded medicine.