Barney,
I think the answer might go something like this:
You've used the figure for Relative Risk to calculate the "apparent"
incidence of oedema in patients given bicarb. Thus the incidence is 1 x 4 =
4%. But the Number Needed to Treat (Harm) is the inverse of the absolute
risk reduction. In this cases the absolute risk reduction is 4% - 1% = 3%.
So the NNT is now 3/100 = 33. In other words, of 33 people given bicarb,
only 1 will experience the outcome (oedema). One would have to make a
judgement whether this is a risk worth taking - I think it is when the
outcome of untreated acidosis is likely to be death!
Most importantly though, your relative risk has a wide confidence interval.
I would recalculate the above for both ends of the limit.
Also, I wonder if your RR of 4 comes from the multiregression analysis? I'd
feel more comfortable if the RR was calculated from the raw data excluding
all the other variables - this would give a better "real life" impression on
the effect of sodium bicarb.
To answer your last point, I think you're correct in saying bicarb is a
marker for severe disease. Its administration is probably a proxy for the
degree of acidosis i.e. those with severe acidosis get bicarb - those with
mild acidosis don't.
Hope this helps
John
Mr John P. Hampson
Teacher Practitioner Pharmacist
Gwenfro Academic Unit
Croesnewydd Road
Wrexham LL13 7YP
United Kingdom
Tel. 01978 727407 or
01978 291100 bleep 5881
Fax 01978 727167
E Mail [log in to unmask]
> ----------
> From: Barney Eskin[SMTP:[log in to unmask]]
> Reply To: Barney Eskin
> Sent: 20 February 2001 21:31
> To: [log in to unmask]
> Subject: Cerebral Edema and DKA
>
> I just finished reviewing the paper by Glaser et al that appeared in the
> New England Journal of Medicine (264, 264-9, 2001) on risk factors for
> cerebral edema in children with diabetic ketoacidosis (DKA). This is a
> multicentered case-control study involving a total of 6977 children with
> DKA of whom 61 developed cerebral edema. Of these, 13 developed permanent
> neurologic sequelae and 13 died. Only 2 children without cerebral edema
> died.
>
> In comparing those with cerebral edema with those without, using logistic
> regression, there appeared to be a lower pCO2 and higher BUN in those
> children with cerebral edema. Of the therapeutic variables, only
> administration of bicarbonate was associated with cerebral edema (relative
> risk = 4.2, 95% confidence interval 1.5-12.1). Use of insulin bolus or
> rate of administration of insulin and rate of IV fluid infusion were not
> associated with an increased risk of developing cerebral edema.
>
> Since the relative risk of giving bicarb for cerebral edema is 4 and the
> incidence of cerebral edema is about 1% (61/6977), does this mean that the
> number needed to harm is about 1 divided by (1% x 4) = 25?
>
> Does this mean that bicarb should never be given to children in DKA? Or
> is giving bicarb merely a marker for severity of disease and therefore the
> risk of developing cerebral edema?
>
> Barney Eskin
> Morristown Memorial Hospital
>
|