Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

NSAIDs, COXIBs, aspirin, and GI bleeds


Guidance from various sources has suggested that any protective effect that coxibs have over non-selective NSAIDs is lost in the presence of low dose aspirin. This was based mainly on a meta-analysis of endoscopy studies with few events. A later meta-analysis showed that coxibs plus low dose aspirin produced fewer ulcers than NSAIDs plus low dose aspirin (Figure 1) [1]. What we have lacked is any evidence with real clinical outcomes. A new and very large observational study [2] fills the gap.

Figure 1: Endoscopic ulcers in pooled analyses of studies with celecoxib, valdecoxib, and rofecoxib (NSAID = any NSAID; coxib = as in the box; aspirin = low dose aspirin)


This was a retrospective cohort study based on prescriptions and hospital admission. It covered all persons aged 65 years or older using outpatient and inpatient services from a database with 91% coverage in Quebec. Those filling a prescription for celecoxib or nonselective (ns)-NSAID over 2.5 years ending in 2002 were included in the cohort if they had one year of prescription data and complete coverage for inpatient services. Exclusions were those with prescriptions for both classes of drugs.

The database had comprehensive details of demographic variables, as well as diagnoses and prescription medicines, so that exposure to the drugs of interest could be evaluated. These included gastroprotective agents.

The main outcome was hospital admission for gastrointestinal perforation or bleeding, obtained from hospital discharge summaries using coded information. Such an admission occurring during an exposure period was attributed to current drug therapy. Where there was overlapping therapy, it was attributed to the drug prescribed last.

A number of analyses were performed for relationships between drug therapies and hospital admission. The main interest was the rates of hospital admission with celecoxib and ns-NSAID, without and with aspirin. Crude results were adjusted for various confounders in the statistical analysis.


The total population in the cohort was 332,491, of whom 71,790 received low-dose aspirin. There were about 700 admissions in almost 70,000 patient years of observation.

Compared with patients taking ns-NSAIDs, those taking celecoxib were more likely to be women, older, and with osteoarthritis. Those taking aspirin had more cardiovascular disease or hypertension than those who did not.

Despite all patients having at least one gastrointestinal risk factor by virtue of age, only about a quarter were receiving gastroprotective therapy at the time of an event, and fewer than a third had a history of having received any such therapy. Higher rates of hospital admission for gastrointestinal perforation or bleeding were seen with older age, in anaemic patients, in those with a prior admission, and with anticoagulant use.

The use of proton pump inhibitor at baseline was associated with a significantly reduced rate of gastrointestinal bleeding compared with non-use (Figure 2). The use of histamine-2 antagonists or misoprostol was not associated with any protection.

Figure 2: Gastrointestinal bleeding with protective agents plus ns-NSAID vs nonuse of protective agent

Compared with ns-NSAIDs, celecoxib was associated with a substantial reduction in admission for gastrointestinal bleeding (Figure 3). Use of low dose aspirin with ns-NSAID and celecoxib substantially increased the rate of hospital admission, but low dose aspirin with celecoxib was better than low dose aspirin plus ns-NSAID, and about the same as ns-NSAID in the absence of low dose aspirin.

Figure 3: Gastrointestinal bleeding with ns-NSAIDs and celecoxib, without and with low dose aspirin


It really is nice when things come together, and when evidence from randomised trials produces the same answer as evidence from observational studies, especially when both fulfil criteria of quality, validity, and size. The surrogate endpoint of endoscopic ulcers (Figure 1) said that coxib plus aspirin was better than ns-NSAID plus aspirin and about the same as ns-NSAID alone, exactly the same as the findings on hospital admission for gastrointestinal bleeding.

In addition, we have confirmation that proton pump inhibitors are protective for actual bleeding events, again much as the surrogate endpoint of endoscopic ulcers showed. Additionally, we can dismiss histamine antagonists, confirming what careful reading of the surrogate endpoints used in randomised trials showed.

All in all, it makes for some substantial rethinking of guidance. Celecoxib plus aspirin is better than ns-NSAID plus aspirin. It would be hard to argue to the contrary.


  1. RA Moore et al. Tolerability and adverse events in clinical trials of celecoxib in osteoarthritis and rheumatoid arthritis: systematic review and meta-analysis of information from clinical trial reports. Arthritis Research & Therapy 2005 7:R644-R665.
  2. E Rahme et al. Hospitalization for gastrointestinal bleeding associated with non-steroidal anti-inflammatory drugs among elderly patients using low-dose aspirin: a retrospective cohort study. Rheumatology 2007 46: 265-272.

previous or next story