Mindstretcher: Making Decisions for Guidelines
This
short piece is by way of being a reminder that making decisions about evidence
is not always easy. You are going to be presented with some evidence about an
important topic, drawn from a Cochrane review. You will have to choose how you
word this in your guidelines.
Choices
There
are three choices (A, B, or C) for a medicine that can reduce the likelihood of
a serious, potentially fatal, event. Randomised trials have been conducted on
identical patient groups, over roughly the same period of time, and using the
same outcome.
Results
The
results are shown in Table 1, and in Figures 1-3.
Table 1: Results for meta-analyses of trials of A, B, and C
(95% CI) |
(95%CI) |
|||||
Figures 1-3: Individual trial results with (from top) A, B, and C
- Medicine A has the smallest number of patients, but a low NNT.
- Medicine B has more patients, and though the NNT is higher than the other two treatments, the proportion of patients with a bad event is lower, at 10%, than the other two treatments.
- Medicine C has the largest number of patients, four times more than medicine A, with equivalent results.
What
do you use to choose? Here are some thoughts.
Medicine
A had too few patients to be sure of the result. The problem with B is that
while it produced the best result in terms of the lowest number of events, the
rate with placebo was also low, yielding higher (worse) relative risk and NNT.
Is the low rate with intervention a reflection of a different population
(probably not), chance (probably not), or some other, unknown, feature of these
particular studies (though some studies with other treatments had this
characteristic; look at Figure 3)? Is medicine C the only safe choice?
Exposure
These
data are for total endoscopic ulcers in trials lasting 3-12 months, from a
Cochrane review updated in 2004 [1]. The treatments were:
A
– High dose H2A (equivalent to at least 300 mg ranitidine twice daily)
B
– Low dose H2A (equivalent to 150 mg ranitidine twice daily)
C
- PPI
Comment
Most
guidelines suggest using PPI or high dose H2A. To do so, they do not rely on
these placebo-controlled studies alone. Nor should they, because the amount of
information is limited, and these are, after all, at best surrogate measures
for much more complicated events.
This
is a case where a single direct comparison can really help. For instance, a
large study compared omeprazole 20 mg daily with ranitidine 150 mg twice daily
in 425 patients, and found omeprazole much superior (6% endoscopic ulcers,
compared with 21% with ranitidine). For high dose H2A we do not even have that
luxury, other than a comparison of two doses of nizatidine, with just eight
events.
Yet
the advice to use double dose histamine antagonists seems to be almost
universal. It is curious that so much can be made of so little.
Reference:
- A Rostrom et al. Prevention of NSAID-induced gastroduodenal ulcers. Cochrane Database of Systematic Reviews, issue 1, 2006.