171 Ashley Ave.
Charleston, SC 29425
843-792-1414
800-424-MUSC
|
 |
November 2006
Evidence-Based Tip
Understanding the Numbers: How to Interpret Statistics in a Study
Authors of studies can measure their results in several ways. To know whether or not the results are important enough to either change or reinforce
our current standard of care, we need to be able to understand those numbers. In a therapy question, those numbers should be able to tell us how
large the treatment effect was, and how precise the effect was.
How large was the treatment effect?
There are three important statistics that help us to understand how effective our intervention was in comparison to either placebo or the alternate
treatment, relative to the outcome we selected.
- Relative Risk is a ratio, not a proportion. It tells us whether the outcome occurs more or less frequently in the group that used
our treatment compared to the comparison group. It is derived by comparing the percentage of people experiencing the outcome in the intervention
group to the percentage of people experiencing the outcome in comparison group.
- Risk Reduction can be relative or absolute. Relative Risk Reductionis a larger number than Absolute Risk Reduction. Both are expressed
as a proportion, and the greater the percentage, the larger the effect of the treatment. For questions of harm, one can use Relative Risk Increase
and/or the Absolute Risk Increase.
- Number Needed to Treat (NNT) is a number derived from the Absolute Risk Reduction, and tells how many patients would needed to be treated
before on would receive the benefit of the treatment. A NNT of five would mean that five patients would have to be treated in order for one patione
to avoid a bad outcome. Likewise, Number Needed to Harm (NNH), is the number of patients who would need to be treated before one patient experienced
the harmful side effect.
How precise was the treatment effect?
Conducting a study is like having a freeze-frame in the total moving picture of how well our intervention works. Did we get enough to see the big picture?
The Confidence Interval tells us the range of what we would expect our results to be if we did this test again and again. If the lower end of the range is
big enough to show our treatment made a positive difference, then we have enough people in our study to get that big picture. How likely is it that we would
get the same results if we did this study again? Most studies are reported with a 95 percent confidence interval. This means that 95 percent of the time,
when we repeated this experiment, the result would fall within our range.
Often, these numbers are given and explained in the report of study, and it is not necessary to do the calculations yourself. However, being aware of what
these numbers mean can help you in your critical appraisal of treatment claims.
|
|
Print Version