Article type
Year
Abstract
Background: Non-steroidal anti-inflammatory drugs (NSAIDs) remain amongst the most widely used drugs, and one of the most important causes of serious adverse drug effects. The development and widespread use of selective COX-2 inhibitors offers increased safety for the upper gastrointestinal GI tract, but possibly not other body systems. Several large randomized trials of the GI safety of COX-2 inhibitors have been published, and there are many large observational studies of upper GI complications (UGIC) involving NSAIDs. The purpose of this systematic review was to integrate the information from these disparate sources.
Methods and Results: We searched the following electronic data-bases: Medline, EMBASE, Cochrane Library, HealthSTAR, CINAHL, Iowa Drug information service (IDIS) for the period 1985 to 2003, and hand-searched key journals (including SCRIP pharmaceutical journal), abstracts from scientific meetings and bibliographies of relevant papers. We selected all large controlled studies (randomised and non-randomised) involving NSAIDs or COX-2 inhibitors (incl comparisons of these drugs) and reported serious UGIC (hemorrhage, perforation, stenosis) in users and non-users of the drugs. We retrieved thirty-six case-control studies that enrolled over 24000 cases and 115000 controls, 14 controlled cohort studies that included over 600,000 exposed and 1.2 million non-exposed subjects (with around 4500 UGIC) and 8 randomised controlled trials (or meta-analyses of other smaller RCTs) that recruited over 50,000 exposed subjects, 30,000 controls, and captured over 250 UGIC. Overall class-specific pooled relative risks for NSAIDs were 4.1 (95% CI 3.5, 4.9) from case-control studies and 2.8 (2.0, 3.9) from cohort studies. RRs for individual NSAIDs ranged from 2.2 (1.7, 2.9) for ibuprofen and 3.9 (3.2 4.7) for diclofenac, to 7.4 (5.8, 9.4) for piroxicam. Individual drug-specific RRs showed clearcut dose responses. The pooled relative risk for serious GI complications with COX-2 inhibitors compared with NSAIDs, estimated from 8 RCTs (or meta-analyses), was 0.53 (0.36, 0.78). This estimate was not changed materially by inclusion of the controversial 12 month (and censored) data from the CLASS study, and was similar to the pooled RR from 2 cohort studies comparing COX-2 inhibitors with NSAIDs (RR 0.40; 0.09, 1.71). The latter is an unadjusted estimate and does not take account of channeling of COX-2 inhibitors to high risk patients. This confounding by indication also compromised attempts to estimate a RR for COX-2 inhibitors compared with no therapy from the observational data, and it is unclear if these drugs cause any increase in the risk of UGI complications. Conclusions: COX-2 inhibitors cause fewer serious upper gastrointestinal complications than NSAIDs. The variation in risk between individual NSAIDs appears to be as great as the difference in the average risks of NSAIDs and COX-2 inhibitors. The risks of these drugs should be viewed as being on a continuum, rather than as bimodal, when making individual treatment choices.
Acknowledgment: This work was supported by a grant from Wyeth Consumer Healthcare, which manufactures brands of ibuprofen
Methods and Results: We searched the following electronic data-bases: Medline, EMBASE, Cochrane Library, HealthSTAR, CINAHL, Iowa Drug information service (IDIS) for the period 1985 to 2003, and hand-searched key journals (including SCRIP pharmaceutical journal), abstracts from scientific meetings and bibliographies of relevant papers. We selected all large controlled studies (randomised and non-randomised) involving NSAIDs or COX-2 inhibitors (incl comparisons of these drugs) and reported serious UGIC (hemorrhage, perforation, stenosis) in users and non-users of the drugs. We retrieved thirty-six case-control studies that enrolled over 24000 cases and 115000 controls, 14 controlled cohort studies that included over 600,000 exposed and 1.2 million non-exposed subjects (with around 4500 UGIC) and 8 randomised controlled trials (or meta-analyses of other smaller RCTs) that recruited over 50,000 exposed subjects, 30,000 controls, and captured over 250 UGIC. Overall class-specific pooled relative risks for NSAIDs were 4.1 (95% CI 3.5, 4.9) from case-control studies and 2.8 (2.0, 3.9) from cohort studies. RRs for individual NSAIDs ranged from 2.2 (1.7, 2.9) for ibuprofen and 3.9 (3.2 4.7) for diclofenac, to 7.4 (5.8, 9.4) for piroxicam. Individual drug-specific RRs showed clearcut dose responses. The pooled relative risk for serious GI complications with COX-2 inhibitors compared with NSAIDs, estimated from 8 RCTs (or meta-analyses), was 0.53 (0.36, 0.78). This estimate was not changed materially by inclusion of the controversial 12 month (and censored) data from the CLASS study, and was similar to the pooled RR from 2 cohort studies comparing COX-2 inhibitors with NSAIDs (RR 0.40; 0.09, 1.71). The latter is an unadjusted estimate and does not take account of channeling of COX-2 inhibitors to high risk patients. This confounding by indication also compromised attempts to estimate a RR for COX-2 inhibitors compared with no therapy from the observational data, and it is unclear if these drugs cause any increase in the risk of UGI complications. Conclusions: COX-2 inhibitors cause fewer serious upper gastrointestinal complications than NSAIDs. The variation in risk between individual NSAIDs appears to be as great as the difference in the average risks of NSAIDs and COX-2 inhibitors. The risks of these drugs should be viewed as being on a continuum, rather than as bimodal, when making individual treatment choices.
Acknowledgment: This work was supported by a grant from Wyeth Consumer Healthcare, which manufactures brands of ibuprofen