Skip navigation

COXIBs and GI ulcers: real world risk reduction


Clinical bottom line

Despite confounding by indication, COXIBs produced lower rates of upper gastrointestinal ulceration than did NSAIDs.


F Wolfe et al. Gastroprotective therapy and risk of gastrointestinal ulcers: risk reduction by COX-2 therapy. Journal of Rheumatology 2002 29: 467-473.


Participants ere enrolled in the National Data Bank for Rheumatic Diseases, and comprised 8,457 patients, most of whom had rheumatoid arthritis and the remainder with fibromyalgia or osteoarthritis. The survey period covered the first six months of 1999, and followed the introduction of COXIBs.

Patients list all drugs used in the study period, with dose, start and stop dates and adverse effects attributed to each drug. Upper gastrointestinal ulcers reported were subjected to validation in which supporting medical records were obtained.

GI therapy that followed development of ulcers was excluded, and combinations of diclofenac and misoprostol were not included. Analysis was by means of logistic regression.


The average age was 60 years, and 80% were women. There was a lifetime history of 17% with upper GI ulcers and 48% with any GI symptoms; 40% used steroids.

The 8,457patients took 12,177 courses of therapy, and analysis was by course of therapy. The most commonly used therapies were celecoxib, ibuprofen, naproxen and nabumetone, which together made up two thirds of courses used. The breakdown of changes in therapy is shown in Table 1:

Table 1: Therapy switching over 12,177 courses

Pattern of use




Two sequentially


Three sequentially


Four or more sequentially


The distribution of therapy with any GPA medication and with proton pump inhibitor is shown in Table 2.

Table 2: Distribution of NSAID and COXIB therapy and GPA


Any GPA (%)

PPI (%)

NSAID - no protective drug



NSAID + protective drug



COXIB - no protective drug



COXIB + protective drug



Rates of ulceration were highest in patients with NSAID + PPI and lowest in those with COXIB without a PPI. Patients taking COXIB + PPI also had a higher ulceration rate than those not taking a PPI.

Overall there was a two-fold lower rate of upper GI ulcers with COXIBS than with NSAIDs (odds ratio 0.45; 95% CI 0.3 to 0.7). In patients using concomitant proton pump inhibitors the rate was decreased four-fold (odds ratio 0.26; 0.15 to 0.44).


We do not know the number of ulcers, and it may well have been small. Caution rules, therefore. The indication from this complicated paper is that use of gastroprotective agent and especially proton pump inhibitor is a reflection of some underlying risk attributable to the patient, the only way to explain the apparently inexplicable - a higher rate of ulceration with patients with NSAID or COXIB plus PPI compared with those not taking a PPI. By this time we can be reasonably sure that PPIs do not cause ulceration, and what we have is something called "confounding by indication".

With or without PPI, though, rates of ulcers were a half or a quarter with COXIBs than those found for NSAIDs. These results mimic the results from randomised trials.