Arnold Kling  

Presenting Health Statistics

Democracy's Vices... Behavioral Vaccine...

Darshak Sanghavi writes

A 31 percent reduction in heart attacks, after all, seems impressive. Yet this pervasive way of describing clinical trials in medical journals—focusing on the "relative risk," in this case of heart attack—powerfully exaggerates the benefits of drugs and other invasive therapies...

There's another instructive way to consider the numbers. Suppose that 100 people with high cholesterol levels took statins. Of them, 93 wouldn't have had heart attacks anyway. Five people have heart attacks despite taking Pravachol. Only the remaining two out of the original 100 avoided a heart attack by taking the daily pills. In the end, 100 people needed to be treated to avoid two heart attacks during the study period—so, the number of people who must get the treatment for a single person to benefit is 50. This is known as the "number needed to treat."

Fair enough. But at least with clinical trials, we can calculate the number needed to treat (NNT), even if it is not the number that gets reported. I respectfully submit that many medical protocols, including referrals for diagnostic imaging and for seeing a specialist, have much worse NNT's, but nobody is tracking those numbers. I'll bet that the number of people you need to send for an MRI when they have back pain in order for a single person to benefit is in the hundreds, if not the thousands. I'll bet that the number of people with microscopic hematuria that you need to send for cystoscopy and IVP in order for a single person to benefit is also quite high.

By the way, Nortin Hadler, in his book The Last Well Person, is also not happy with the relative risk measure of effectiveness.

Thanks to Trent McBride (email) for the pointer).

Comments and Sharing

COMMENTS (3 to date)
bill writes:

I bet it is. This is a little off topic, but something I've wondered about and noticed about your posts regarding the lack of costs incentives to individual consumers under our current system of health insurers paying for costs:

The health insurers are paying for things like this, they, if anyone, would hypothetically be able to keep costs to a minimum by paying very little or nothing for unnesseccary tests. Meanwhile, because it is the medical providers who have the knowledge of who needs what procedures, and also the medical providers who tell people what treatement they need, does it not make more sense that medical providers (in part perhaps in fear of law suits, but also because it means more money for them) are the ones overtreating and driving up medical costs, at least, more so than a normal insured consumer without normal costs incentives is? I mean, while some people like consuming medical treatment and getting fussed over and seeing people worse off than themsleves, I think the majority of people love to hear that they don't need anymore treatment.

I am confused as to why costs aren't kept lower by the heatlh insurers though, auto insurers definately aren't paying too much or for unneccessary auto repairs. Perhaps because of too little competition in the medical provision field (due to licensing and regulation and all the work it takes to be a doctor, P.A., or nurse)?

Dr. T writes:

I agree that using percent change in the relative risk is misleading. Unfortunately, many physicians remember the big number and fail to calculate the actual clinical impact (which often is negligible). What irritates me more is when medical journal editors and expert commenters echo the misleading statistics and support high cost interventions that have little benefit.

justin writes:

It does sound like the NNT is a good statistic to use. But for drugs that one can take routinely for a long time, like statins, it seems like the story is a little more complicated. The article says the NNT from a five-year study of statins was 50. What if you take a statin for 20 years? It seems that you would get extra chances to be the lucky 1-in-50 who benefits. So maybe something like the "chance of benefit per year" would be appropriate for these types of treatment.

But some of the benefits of statins might be cumulative. After the first five year period, your indicators for heart disease (LDL, HDL, etc.) may be more favorable if you have taken the drug. Would this mean you have a lower risk for heart disease in the following five years? Would it mean the risk has become low enough that there is a lower benefit to taking the drug for five more years?

Obviously that sort of complication can not occur with a diagnostic like MRI. I don't think that an MRI today can benefit you 5 years from now.

Comments for this entry have been closed
Return to top