Current issues of ACP Journal Club are published in Annals of Internal Medicine


Meta-analysis: Low-dose ibuprofen has the lowest gastrointestinal risk of any NSAID

ACP J Club. 1996 Nov-Dec;125:78. doi:10.7326/ACPJC-1996-125-3-078

Source Citation

Henry D, Lim LL, Rodriguez LA, et al. Variability in risk of gastrointestinal complications with individual non-steroidal anti-inflammatory drugs: results of a collaborative meta-analysis. BMJ. 1996 Jun 22;312:1563-6.



To determine the relative risks for serious gastrointestinal (GI) complications associated with specific nonsteroidal anti-inflammatory drugs (NSAIDs) at various doses.

Data sources

Studies were identified using MEDLINE (1985 to 1994), bibliographies of relevant review articles, and by contacting authors.

Study selection

Case-control and cohort studies were selected if an association was found between the use of NSAIDs by adults living in the community and serious GI complications that required hospitalization for hemorrhage or perforation.

Data extraction

Data were extracted on individual NSAIDs and dose. The pooled relative risk (RR) for each NSAID was calculated using the unadjusted RR with an RR of 1.0 assigned to ibuprofen. In addition, a weighted summary rank of the RR was calculated for each NSAID. Individual authors were funded to meet to discuss data extraction, and they supplied missing information as needed.

Main results

Fenoprofen, diflunisal, tolmetin, and azapropazone were each included in 2 studies; all other NSAIDs were included in 5 to 11 studies. The pooled RRs for serious GI complications for each NSAID are listed in the table in increasing order of RR with the weighted mean rank. Higher doses increased the risk for serious GI complications for the 3 drugs that had sufficient data to study dose-response relations (ibuprofen: low-dose RR 1, high-dose RR 4.2 [CI 1.8 to 9.8]; naproxen: low-dose RR 3.7 [CI 1.7 to 7.7], high-dose RR 6.0 [CI 3.0 to 12.2]; indomethacin: low-dose RR 3.0 [CI 2.2 to 4.2], high-dose RR 7.0 [CI 4.4 to 11.2]).


Of the 12 nonsteroidal anti-inflammatory drugs, low-dose ibuprofen has the lowest risk for serious gastrointestinal complications requiring hospitalization. The individual drug and its dose both affect the risk for serious gastrointestinal complications.

Source of funding: Boots Australia Pty Ltd.

For article reprint: Dr. D. Henry, Discipline of Clinical Pharmacology, Faculty of Medicine and Health Sciences, Level 5, Clinical Sciences Building, Mater Hospital, Edith Street, Waratah, New South Wales 2298, Australia. FAX 61-49-602-088.

NSAID RR (95% CI) Mean Rank
Ibuprofen 1.0 1.0
Fenoprofen 1.6 (1.0 to 2.5) 3.5
Aspirin 1.6 (1.3 to 2.0) 4.8
Diclofenac 1.8 (1.4 to 2.3) 2.3
Sulindac 2.1 (1.6 to 2.7) 6.0
Diflunisal 2.2 (1.2 to 4.1) 3.5
Naproxen 2.2 (1.7 to 2.9) 7.0
Indomethacin 2.4 ( 1.9 to 3.1) 8.0
Tolmetin 3.0 (1.8 to 4.9) 11.0
Piroxicam 3.8 (2.7 to 5.2) 9.0
Ketoprofen 4.2 (2.7 to 6.4) 10.3
Azapropazone 9.2 (4.0 to 21.) 11.7


High-dose famotidine prevented NSAID-induced ulcers in arthritis

The meta-analysis by Henry and colleagues chiefly addresses the severe upper GI complications associated with NSAIDs; such complications are by far the most common serious consequences of therapy with these drugs. The meta-analysis assigns a rank order that is probably valid for the more widely used NSAIDs, but points out that all drugs with a long enough history of use (which excludes several popular drugs introduced in the past few years) have GI consequences. The narrow CIs confirm that the rankings are probably accurate, with most NSAIDs clustered in a relatively tight intermediate grouping. The recommendation that is derived from the ranking is to first use the NSAIDs that are least often associated with upper GI problems and then, based on the patient's need and response, advance to others that have higher adverse effect ranks.

Ibuprofen is the standard NSAID but only in doses to 2400 mg/d; at higher doses, it ranks much lower. The reason for administration may differ for the various drugs; some that are perceived as more analgesic are used for milder conditions, and others that are more anti-inflammatory are given to patients with greater need and possibly greater risk for complications. Case-control and cohort studies cannot address the distinctions in patients' choices or in the utility of the drugs as perceived by physicians.

The study by Taha and colleagues presents welcome news because it discusses a way to reduce upper GI risk for patients through co-treatment with famotidine, a histamine H2-receptor antagonist. The population studied, however, differs from the epidemiologic meta-analysis of the study by Henry and colleagues, which dealt with overt clinical cases. Most previous studies have failed to find a reliable connection between endoscopically discovered problems and overt upper GI diseases in NSAID users, but most of the upper GI injuries in the study by Taha and colleagues were discovered by endoscopy. Half of the study patients harbored Helicobacter pylori—the increased risk for cancer in these patients has recently been analyzed (1), and the association between ulcers and cancer of the stomach seems to be confirmed, at least for gastric ulcers. For duodenal ulcers (for which histamine H2-blockers are the best preventive agents), this relation does not exist. This raises the possibility that successful ablation of symptoms from gastric erosion might delay discovery of a problem that is even worse than erosive GI disease. In addition, there is the risk of polypharmacy, with its potential complications and costs. In his recent commentary, Carson (2) cited a study that suggested that acetaminophen, a non-NSAID that does not cause upper GI complications, provided satisfactory treatment for osteoarthritis (3). This is true for analgesia, but to treat inflammation only NSAIDs offer the appropriate effects. The controversy about the difference in outcome is not settled. Several of Carson's conclusions are addressed in these 2 studies: The risk for GI complications remains low, the relation between NSAIDs and upper GI disease is real, and the relative toxicity of individual NSAIDs can be identified. We still need to better understand for whom, how, and why each of the NSAIDs was prescribed and whether the long-term addition of a presumed ulcer prophylaxis will result in safer, more effective, and less expensive treatment.

George E. Ehrlich, MD
University of Pennsylvania and New York UniversityPhiladelphia, Pennsylvania, USA


1. Hansson LE, Nyren O, Hsing AW, et al. N Engl J Med. 1996;335:242-9.

2. Carson JL.Review: Nonaspirin NSAIDs cause adverse gastrointestinal events. ACP J Club. 1992 Mar-Apr;116:60. Comment on: Gabriel SE, Jaakkimainen L, Bombardier C. Ann Intern Med. 1991;115:787-96.

3. Bradley JD, Brandt KD, Katz BP, Kalsinski LA, Ryan SI. N Engl J Med. 1991;325:87-91.