Observational study of upper gastrointestinal haemorrhage in elderly patients given selective cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs

In the hierarchy of evidence, well designed RCTs provide more
reliable answers than observational studies(1). However, you have
published an observational cohort study by Mamdani and colleagues(2), in
which the authors conclude that there are lower rates of upper GI
haemorrhage (UGIH) with selective COX 2 inhibitors than with non-selective
NSAIDs.

As you are aware, there have been two large RCTs comparing coxibs
with traditional NSAIDs, CLASS(3) and VIGOR(4). CLASS compared celecoxib
with ibuprofen and diclofenac, and although the published 6-month data
suggests that celecoxib is slightly safer in terms of serious upper GI
events, the 12-month data available on the FDA website puts this
conclusion into serious doubt(5).

In VIGOR, rofecoxib was compared with naproxen and although in terms
of GI safety it does appear safer, its overall safety compared to naproxen
is questioned(6).

So, what does this observational study add? Firstly, the authors
present the data for UGIH as both unadjusted and adjusted risk ratios. The
adjusted ratio for celecoxib is 1.0, the same as the reference (community
controls), and they therefore do not allocate a NNH for celecoxib.
However, the unadjusted risk ratio for celecoxib is 1.7 (95% CI 1.1 to
2.6), showing increased risk.

Using the data presented in Table 2, the incidence of admissions for
UGIH calculate as follows:

No NSAID - 0.082%, NSAID - 0.315%, Diclo + miso - 0.256%, Rofecoxib -
0.295%, Celecoxib - 0.169%.

The graph on page 626, which presents percentages of patients
hospitalised against time, appears to use different figures for rofecoxib
and celecoxib than calculated from Table 2. The line for celecoxib is then
identical to that for control and that for rofecoxib is much lower than
that for diclofenac + misoprostol. This is not supported by the data from
the table.

The absolute difference in incidence of admissions for UGIH for
celcoxib over non-selective NSAIDs is 0.146%. This calculates to an NNT of
685, which is not exactly impressive and arguably not cost-effective.

Subjects were 66 years of age or more, with a mean age around 76. Yet
the rate of admissions for UGIH in NSAID users was only 0.315%, compared
to 0.082% in non-users (NNH = 429). This questions the NICE decision that
age as a risk factor is 65 and arguably it should be around 75.

We are currently unsure of the GI and CV comparative toxicity of
coxibs over traditional NSAIDs and this paper adds more to the controversy
rather than helping to resolve it. Perhaps pressure should be put on to
NICE to bring forward their review of COX 2 selective NSAIDs?

1.Eccles M et al. BMJ 1998; 316:1232-5

2.Mamdani M et al. BMJ 2002; 325: 624-9

3.Silverstein FE et al; JAMA 2000; 284:1247-55

4.Bombardier C et al. N Engl J Med 2000; 343:1520-8

5.Juni P et al. BMJ 2002; 324: 1287-8

6.Budenholzer BR. BMJ 2002; 325: 161

No competing financial interests -- I am an NHS employee.