Letters Statistics for health

Let’s start with reporting

BMJ 2009; 339 doi: https://doi.org/10.1136/bmj.b3035 (Published 28 July 2009) Cite this as: BMJ 2009;339:b3035

This doesn't make sense

This letter provides misleading advice. It doesn't make sense to
apply a number
needed to treat (NNT) from a meta-analysis that includes trials with
widely
differing populations that have widely differing underlying likelihoods of
risk.
The NNT is based on the absolute risk reduction, not a relative one (which
is
more constant and therefore makes sense for applying in a meta-analysis).
If
the underlying risk varies widely, say from primary prevention trials
(such as the
JUPITER trial) to diabetes patients (such as the HPS), then the NNT
provides a
spurious estimate of the number of people needed to be treated to prevent
an
event. Absolute risks really aren't useful in most meta-analyses.

Competing interests:
None declared

Competing interests: No competing interests

02 August 2009
Marko Tostad
Epidemiologist
NY