I read with interest the article by Rahme et al (1) reporting the results of a retrospective cohort study that examined the rates of gastrointestinal (GI) adverse events in elderly patients with arthritis who were treated with acetaminophen or conventional nonsteroidal antiinflammatory drugs (NSAIDs). The rate of dyspepsia symptoms in the acetaminophen cohort appeared to be dose dependent (1). Thus, patients receiving a daily dose of acetaminophen exceeding 2,600 mg were found to have rates of GI events comparable with those observed in the NSAID cohort, even after adjustment for GI risk susceptibility (1).
I believe these findings conflict with those reported by the same group in an earlier publication (2). Using the same Government of Quebec's health insurance database, the authors obtained medical, pharmaceutical, and demographic records of persons ages 65 years and older who, between 1993 and 1997, had either a nonselective NSAID or an acetaminophen prescription dispensed. The patients were followed up for 2 years. The adjusted risk of a GI adverse event for patients in the NSAID cohort was more than twice that for patients in the acetaminophen cohort (odds ratio 2.48, 95% confidence interval [95% CI] 2.06–3.00). Accordingly, the mean direct cost of GI events for a day of conventional NSAID therapy was $0.94 (Canadian) (95% CI 0.28–1.65) in the NSAID cohort, compared with $0.10 (95% CI 0.06–0.12) for a day of acetaminophen therapy in the acetaminophen cohort (2).
The main limitations of the study by Rahme et al (1) were defined in the excellent accompanying editorial (3). I would like to add that results of case–control studies may vary according to the data collection protocols used. A meta-analysis of individual patient data from 3 case–control studies using incident cases and information obtained by direct questioning of all patients and controls showed that acetaminophen was not associated with upper GI bleeding at any dosage (4). In contrast, the authors of a nested case–control study using computerized prescription data concluded that patients receiving acetaminophen at daily doses greater than 2 gm had an increased risk of upper GI bleeding or perforation similar to that in patients receiving traditional NSAIDs (5). In fact, the notion that acetaminophen exerts serious detrimental effects on the GI mucosa seems hard to accept, because millions of patients, including those at higher risk for GI events, have used this drug. It seems unlikely that such effects have been overlooked for several decades. Moreover, no data indicate that acetaminophen inhibits cyclooxygenase 1 (COX-1) in gastric mucosa (3, 6). Interestingly, a third distinct COX isoenzyme, COX-3, which is a covariant of COX-1, was recently identified in the central nervous system (6). Unlike COX-1 and COX-2, COX-3 was shown to be inhibited by acetaminophen at concentrations achieved at therapeutic dosages (6). Thus, COX-3 could be the biologic target by which acetaminophen exerts its analgesic and possibly antipyretic properties (6).
Finally, I endorse the opinion expressed by Abramson (3), that “the weight of the clinical evidence continues to support the superior overall GI safety profile of acetaminophen compared with nonselective NSAIDs.”