Gastrointestinal bleeding after the introduction of COX 2 inhibitors: ecological study

Introduction

Recent evidence suggests a lower risk of upper gastrointestinal haemorrhage for selective cyclo-oxygenase-2 (COX 2) inhibitors compared with non-selective non-steroidal anti-inflammatory drugs (NSAIDs) at the patient level,1^–3 although COX 2 inhibitors are likely not devoid of gastrointestinal toxicity. At the population level, however, the widespread proliferation of COX 2 inhibitors might lead to an increase in the overall numbers of people exposed to anti-inflammatory drugs with uncertain implications on rates of population-wide gastrointestinal events. We did an ecological study to examine temporal changes in the use of NSAIDs and upper gastrointestinal haemorrhage hospitalisation rates among a population of older individuals after the introduction of COX 2 inhibitors.

Participants, methods, and results

We did a population based cross sectional time series analysis using administrative healthcare databases covering more than 1.3 million residents of Ontario, Canada, aged at least 66 years.4 This population has universal access to hospital care, doctors' services, and prescription drugs on a formulary. The study's timeframe was divided into 15 intervals of six months from 1 September 1994 to 28 February 2002. Rofecoxib and celecoxib were introduced on the provincial drug formulary in April 2000 and meloxicam was introduced in March 2001. The prevalence of use of NSAIDs in each interval was determined by dividing the unique number of individuals dispensed any NSAID (either non-selective NSAIDs or COX 2 inhibitors) by the total number of individuals alive at the beginning of the interval. Similarly, we examined hospitalisation rates for upper gastrointestinal haemorrhage. As secondary endpoints, we examined hospitalisations for myocardial infarction and heart failure. We standardised all rates for age and sex. As supplementary analyses, we also examined changes in the use of gastroprotective agents, oral corticosteroids, prescription aspirin, and warfarin, since these factors may be strongly related to upper gastrointestinal haemorrhage. We used time series analysis involving autoregressive integrated moving average models to evaluate changes over time with the package SAS 8.2 (SAS, Cary, NC).5

The prevalence of use of NSAIDs among Ontario's population of older people increased from 14.0% just before the introduction of COX 2 inhibitors to 19.8% by the end of the observation period (figure; P < 0.01), representing an absolute increase of more than 90 000 additional individuals annually using NSAIDs, entirely attributable to the use of COX 2 inhibitors rather than switching from non-selective NSAIDs to COX 2 inhibitors. The rate of hospitalisation for upper gastrointestinal haemorrhage was decreasing before the introduction of COX 2 inhibitors, but increased from about 15.4 to 17.0 per 10 000 older persons after their introduction (figure; P < 0.01), representing an absolute increase of more than 650 upper gastrointestinal haemorrhage hospitalisations annually. Other than a small but statistically significant increase in the prevalence of gastroprotective agent use, we saw no significant differences in the use of drugs that might affect upper gastrointestinal risk over expected projections. Also, we saw no significant differences in hospitalisation rates for myocardial infarction or heart failure greater than expected projections.

Age and sex standardised prevalence of the use of NSAIDs and hospitalisation rates for upper gastrointestinal haemorrhage over time among elderly people in Ontario

• Download figure
• Open in new tab
• Download powerpoint

Comment

In this population based study a 41% rise in NSAID use, entirely due to increased use of COX 2 inhibitors, was accompanied by a 10% increase in hospitalisation rates for upper gastrointestinal haemorrhage. Although we cannot prove causation, we believe that the striking temporal correlation, biological plausibility, and lack of any other trends that would explain the association strongly suggest that the two events are directly related. Coding practices for hospital admissions for upper gastrointestinal haemorrhage period did not change significantly during our study. However, we could not evaluate whether the potential improvement in population level pain relief offsets the increase in hospitalisations for upper gastrointestinal haemorrhage.

The findings of this study suggest that even if a new drug is associated with lower side effects than previous drugs in its class at the patient level, a marked increase in its use can be associated with an apparently paradoxical adverse impact on the population.