To the Editor:

In a retrospective cohort study, Rahme et al compared the rates of gastrointestinal (GI) events occurring with higher versus lower doses of acetaminophen (1). They concluded that patients receiving higher doses of acetaminophen have higher rates of GI events compared with those receiving lower doses. Their conclusion is in contrast to the large body of well-controlled, prospective data indicating that acetaminophen is unlikely to cause gastric irritation. This discrepancy is likely a result of the severe limitations of their study design.

The authors' conclusion is a biased observation and is confounded by indication. That is, a population that was at higher risk for GI bleeding was prescribed acetaminophen. For example, according to Table 2, several of the common GI risk factors were found statistically significantly more frequently in patients prescribed acetaminophen: prior history of a GI event, older age, greater severity of chronic disease, corticosteroid and anticoagulant use, and use of low-dose aspirin in the previous 3 months. Thus, it is likely that the incidence of GI events in the acetaminophen cohort would have been higher even in the absence of exposure to acetaminophen. The authors acknowledge that patients in the acetaminophen group overall were sicker than those in the nonsteroidal antiinflammatory drug (NSAID) group.

Additionally, Rahme et al conclude that in the high-risk acetaminophen users, the risk of GI events increased with increasing acetaminophen doses. However, this also is probably a reflection of confounding by indication. It is likely that patients who received higher doses of acetaminophen were also those who were older and had higher rates of prior GI events and use of concomitant medications.

The investigators observed that rates of GI events were higher among acetaminophen users than among patients who received NSAIDs. Again, this is confounding due to a higher-risk population in the acetaminophen group (Table 2). Moreover, in the NSAID cohort, prophylaxis with GI protective agents was prescribed at a higher rate (almost a 4:1 ratio) (see Table 3). Thus, the perceived low rate of GI events in the NSAID group was likely artificially lowered by the much higher rate of GI prophylaxis in NSAID users. Gastroprotective agents are also very effective for dyspepsia. Thus, the perceived lower rates of dyspepsia in NSAID users are also likely related to the higher rate of prescriptions for agents used to treat dyspepsia (i.e., GI protective agents) in this group.

According to Rahme et al, GI events were defined by the combination of “GI hospitalization,” ulcers, and dyspepsia. However, the authors indicate that in this study, GI event rates were accounted for mostly by dyspepsia. Thus, the above confounding in the dyspepsia group is the factor most likely responsible for confounding in the reported rates of GI events. Also, as mentioned above, the authors do not state whether patients receiving higher doses of acetaminophen had higher rates of a history of ulcer disease or GI hospitalizations. If the high-dose acetaminophen group did indeed include higher proportions of ulcer patients, those patients would be expected to have higher rates of dyspepsia attributable to their ulcer disease. This is another instance of confounding by indication.

Another limitation is the duration of followup. In this study, the average followup of acetaminophen users was only 13.5 days. Complications of ulcer disease such as GI bleeding commonly recur, particularly within the first several days of the initial bleed. In the first 2 weeks after a GI complication, the observed rates of recurrent complications are more likely a consequence of the residual effect of the presenting ulcer complication rather than a consequence of the acetaminophen that the patients were prescribed to lower the risk of aggravation of their ulcer diathesis.

Finally, the authors suggest several times that acetaminophen has been shown to inhibit cyclooxygenase 1 (COX-1) and therefore might have a pharmacologic basis for gastric toxicity. However, the previous literature has shown that acetaminophen is a weak, isoform-nonspecific COX inhibitor in platelets only (2). Feldman and I have demonstrated that at very high doses, acetaminophen has no effect on COX-1 in the stomach (3).

In addition to pharmacologic data, clinical data indicate that recommended doses of acetaminophen do not cause gastric irritation, gastric erosions, occult or overt GI blood loss, or ulcers (4, 5). In fact, acetaminophen is the recommended pain reliever for use in patients with a history of GI bleeding, because it is unlikely to cause the gastric irritation often associated with aspirin, naproxen sodium, or ibuprofen.

Several case–control studies have established that GI bleeding is a significant risk associated with both regular and occasional use of aspirin or NSAIDs, whereas acetaminophen is not associated with a risk for GI bleeding (6–8). A case–control study evaluating first-time peptic ulcer patients found no significant risk associated with acetaminophen (9). An American College of Gastroenterology survey found that over-the-counter aspirin and NSAIDs were used significantly more often by patients with GI disorders than by controls. However, this was not the case with acetaminophen (10).

In summary, the conclusions of Rahme et al regarding the GI risks of acetaminophen contradict a large body of evidence summarized in this letter, which, collectively, in large part strongly indicates that acetaminophen is not associated with clinical toxicity in the GI tract. The authors' observations regarding acetaminophen are most likely incorrect for reasons of confounding and biases inherent in their study population.

  • 1
    Rahme E, Pettitt D, LeLorier J. Determinants and sequelae associated with utilization of acetaminophen versus traditional nonsteroidal antiinflammatory drugs in an elderly population. Arthritis Rheum 2002; 46: 304654.
  • 2
    Catella-Lawson F, Reilly MP, Kapoor SC, Cucchiara AJ, DeMarco S, Tournier B, et al. Cyclooxygenase inhibitors and the antiplatelet effects of aspirin. N Engl J Med 2001; 345: 180917.
  • 3
    Cryer B, Feldman M. Comparison of effects of celecoxib, rofecoxib, naproxen and acetaminophen on gastric COX inhibition [abstract]. Am J Gastroenterol 2002; 97 Suppl 9: S57.
  • 4
    Hoftiezer JW, O'Laughlin JC, Ivey KJ. Effects of 24 hours of aspirin, Bufferin, paracetamol and placebo on normal human gastroduodenal mucosa. Gut 1982; 23: 6927.
  • 5
    Johnson PC, Driscoll T. Comparison of plain and buffered aspirin with acetaminophen in regard to gastrointestinal bleeding. Curr Ther Res 1981; 30: 7984.
  • 6
    Begaud B, Chaslerie A, Carne X, Bannwarth B, Laporte JR, Sorbette F, et al. Upper gastrointestinal bleeding associated with analgesics and NSAID use: a case-control study [letter]. J Rheumatol 1993; 20: 14434.
  • 7
    Laporte JR, Carne X, Vidal X, Moreno V, Juan J. Upper gastrointestinal bleeding in relation to previous use of analgesics and nonsteroidal anti-inflammatory drugs. Lancet 1991; 337: 859.
  • 8
    Levy M, Miller DR, Kaufman DW, Siskind V, Schwingl P, Rosenberg L, et al. Major upper gastrointestinal tract bleeding: relation to the use of aspirin and other nonnarcotic analgesics. Arch Intern Med 1988; 148: 2815.
  • 9
    McIntosh JH, Fung CS, Berry G, Piper DW. Smoking, nonsteroidal anti-inflammatory drugs and acetaminophen in gastric ulcer: a study of associations and the effects of previous diagnosis on exposure patterns. Am J Epidemiol 1988; 128: 76170.
  • 10
    Peura DA, Lanza FL, Gostout CJ, Foutch PG. The American College of Gastroenterology Bleeding Registry: preliminary findings. Am J Gastroenterol 1997; 92: 9248.

Byron L. Cryer MD*, * University of Texas Southwestern Medical Center and Dallas Veterans Administration Medical Center Dallas, TX.