HONcode Certified

This website is certified by Health On the Net Foundation. Click to verify.

This site complies with the HONcode standard for trustworthy health information: verify here.

Understanding evidence-based medicine in 4 days. Lesson 1: Clinical significance is all about risk

Ami Banerjee
Last edited 17th March 2010

It is often hard to figure out the findings of health research because of jargon and the numbers. However, I reckon most of that research can be understood by anybody with 4 simple concepts. I am going to cover one of these concepts each day using stories from this week’s press relating to health to show how often these numbers appear in the press. Hopefully these 4 keys will allow more people to open the door and to question the numbers we read about in health research.


For over 2000 years, two principles have formed the basis of medical practice: “primum non nocere” (first do no harm) and “succurrere” (do good). If we want to measure “the good” or “the harm” associated with a treatment or an exposure, we have to know how it changes the chance or risk of a disease compared to another treatment or exposure. Chance or risk is usually expressed as a percentage, and tells us about the number of people who develop a disease out of a population.

In absolute terms, this change is simply the difference between the risk associated with the first, or control, treatment and the risk associated with the new treatment. This difference is sometimes called the absolute risk difference. In relative terms, this same change can be expressed as the risk associated with the new treatment divided by the risk associated with the control treatment, known as the relative risk.

In this week’s British Medical Journal, Dutch researchers looked at whether 15 minutes of immobilisation increased the chance of successful pregnancy after artificial insemination . 199 couples received the new treatment (15 minutes immobilisation) and 192 couples had standard treatment (they were allowed to mobilise immediately after insemination- the control group). In the immobilisation group, 54 couples had pregnancies. Therefore the chance or risk of pregnancy was 54/199= 27% in this group. In the control group, 34 out of 192 couples had pregnancies, and so the risk of pregnancy was 34/192=18%.

The absolute risk difference is 27%-18%=9%. In other words, immobilisation increased the risk of pregnancy by 9%, compared with controls. Put another way, the relative risk was 27/18=1.5. This means that compared to standard practice, immobilisation leads to a 50% increased chance of pregnancy after insemination. You might have spotted that a 50% increase sounds a lot more impressive than a 9% increase! Therefore, scientists should either report both absolute and relative risks, so that we can understand the size of the risk, or report absolute risk because it is more useful. As readers, we should also look for these numbers before making any conclusions about harm and good. The terms, “hazard” and “odds” are sometimes used in research, but they are just slightly different measures of chance. The message is still the same: absolute changes caused by a treatment are often smaller than the relative changes.

Lesson 2

Chance and risk are just two

Chance and risk are just two factors, and both of which require an amazing level of competence on the decision maker in order to realize.
Consensus is needed, or near consensus, before action should be undertaken.
Physician level care should be, as much as possible, guided by prevailing theory/evidence, otherwise it's irresponsible.

what matters more

Neat article and couldnt agree more

The reporting of relative effect measures at the expense of absolute measures makes clinical interpretation difficult. The antiplatelet triallist collboaration made this point 15 years ago

'From a medical point of viewpoint what chiefly matters is not the proportional rediction in risk but the absolute reduction in risk'
BMJ 1994; 308(6921):81-106

looking forward to lesson 2

Cheers Carl

Twitter TrustTheEvidence.net


Search the TRIP Database

TRIP Database


Recent Comments