Last year, the UKPDS and HOT studies generated debate because NNTs
calculated by some members of this list were different to those the authors
published. This was mirrored at critical reading seminars held by general
practitioners in my town. Clearly, the two methods of calculating the NNTs
when events are reported as patient years give different results.
As far as I know, the discussion petered out (sorry if I missed a vital step
and am wrong on this). I would like to revive the debate. As I see it, you
can calculate NNTs from either
1. the difference in events per patient years OR
2. the difference in event rates (standard CER and EER) over the average
duration of follow up.
E.g. Take the total CVS major events in the HOT study and NNTs posted by
Atle Klovning on his web site recently. He used method 2. If I have
transcribed his figs correctly, (sorry Atle if I have erred), he gets: NNT
482 (over 3.8 years)
Using method 1, I get: NNT167 (per 1000 patient years)
There is a danger of talking at cross purposes.
I planned to post some contrasts on the Suffolk Primary Care Resource Centre
site (http://suffolk-maag.ac.uk) but have held back in case anyone on this
list feels they have resolved the issue!
Kev Hopayian
Seahills
Aldeburgh Rd
Aldringham
Suffolk
IP16 4QL
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|