Dear Simon,
Most NNT statements require a timeframe, e.g, I have to treat X
hypertensive for 5 years to prevent 1 stroke. If you change the time-frame
then the NNT changes.
So, expressing the zinc result as: treating 6.25 children for 1 year would
prevent 1 LRTI seems a reasonable statement. (or 3.125 for 2 years, or 1
for 6.25 years, are all 6.25 person years).
Were you also concerned about the possibility of multiple events (LRTIs) in
the same person? This makes the stats more difficult, but the point of NNT
is to express the summary estimator in concrete terms (and the multiple
events business only becomes important in calculating the confidence
intervals).
Paul Glasziou
-----
At 12:17 PM 12/10/98 -0500, Simon, Steve, PhD wrote:
>I'm reading a paper for a journal club (Zinc Supplementation Reduces the
>Incidence of Acute Lower Respiratory Infections in Infants and Preschool
>Children: A Double-blind Controlled Trial, Sazawal S et al Pediatrics (July
>19998), 102 (1) 1-5). They mention that the rate of acute lower respiratory
>infections (ALRI) decreased from 0.35 episodes per child per year to 0.19.
>
>If I calculate a number needed to treat (6.25), does this have any meaning?
>It seems to me that this means that if we use zinc supplements, we will see
>one fewer ALRI for every 6.25 patient years. Is this a valid statement? Or
>can I only use the NNT for proportions?
>
>Also, I assume that ALRI is seasonal (I'm not a doctor, so forgive my
>naivete). If so, does this change the interpretation (e.g., 6.25 patient
>years during the such and such a time period)?
>
>Sorry if this is such a basic question. I have a background in Statistics,
>but this NNT stuff is new to me.
>
>Steve Simon, [log in to unmask], Standard Disclaimer.
>STATS - Steve's Attempt to Teach Statistics: http://www.cmh.edu/stats
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|