Dear Madam/Sir:
With the purpose of teaching the academic benefits of the EBM, I did
analyze the following paper "Besarab A et al. The effects of normal as
compared with low hematocrit values in patients with cardiac who are
recieving hemodialysis and epoietin. N Engl J Med 1998;339:584-90."
They examined the risks and benefits of normalizing the hematocrit in
patients with cardiac disease who were undergoing hemodialysis.
The primary end point was the lenght of time death or a first nonfatal
myocardial infarction.
These are the data and I calculated NNH but investing the data. Is this
correct?
Arcus Quickstat (trial version) does not allowed calculate NNH, so I
invested the data.
A)Number needed to treat (is NNH the inverse of NNT?)
Proportion of controls suffering an event = 150/615 = ,243902
Proportion of treated suffering an event = 183/618 = ,296117
With near exact 95,% confidence intervals:
Relative risk = 1,214078 (1,009847 to 1,460831)
Relative risk reduction = -,214078 (-,460831 to -,009847)
Absolute risk reduction = -,052214 (-,101628 to -,002634)
Number needed to treat = -19, (-380, to -10,)
B) Number needed to harm
Proportion of controls suffering an event = 183/618 = ,296117
Proportion of treated suffering an event = 150/615 = ,243902
With near exact 95,% confidence intervals:
Relative risk = ,823671 (,684542 to ,990245)
Relative risk reduction = ,176329 (,009755 to ,315458)
Absolute risk reduction = ,052214 (,002634 to ,101628)
Number needed to treat (harm) = 19, (10, to 380,)
Does this mean that of each 19 patients ( 10 to 380 ) that receive the
experimental treatment 1 harmed it will be?
What do you think about that?
I wait your valuable comment
Many thanks
Arturo Martí-Carvajal
Clinical Epidemiology Unit
Universidad de Carabobo
Venezuela
E-mail:[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|