Background: Use of non‐steroidal anti‐inflammatory drugs (NSAIDs) increases the risk of acute myocardial infarction (AMI), but the magnitude varies across individual NSAIDs and is unknown for some NSAIDs.
Objectives: To estimate the risk of AMI associated with individual NSAIDs.
Methods: Design: Nested case–control study in an adult new NSAID user cohort. Each case was matched to 100 controls on database, sex, age, and event date by using common data model and software for data transformations.
Setting: Six European healthcare datasources: IPCI, PHARMO, NL; SISR, OSSIFF, IT; GePaRD, DE and THIN, UK; 1999–2011.
Exposure: All captured NSAIDs. Dose and duration analyses were conducted in IPCI, PHARMO and THIN.
Outcome: Cases had a coded diagnosis of AMI.
Statistical analysis: Adjusted odds ratios (ORs) were estimated in each datasource for current compared to past use. Pooling was done by a random effects model (ORmeta) and combining datasets on individual level (ORpool).
Results: We matched 79,553 AMI cases to almost 7.5 million controls. Odds ratios were estimated for 28 individual NSAIDs when combining datasets on individual level. The risk was highest for ketorolac (ORmeta 2.06; ORpool 1.80) followed by indometacin (ORmeta 1.47; ORpool 1.51) then etoricoxib, diclofenac and rofecoxib (ORmeta ranged 1.26–1.31 and ORpool 1.28–1.39). Duration of use of ≥90 days and 1–6 days were associated with higher risks for various NSAIDs compared to duration of 7–29 days. Doses >1.2 times DDD for celecoxib, ibuprofen and naproxen showed higher risk estimates than normal (0.8–1.2DDD) doses. Effect modification by sex (diclofenac), age (aceclofenac, naproxen), history of ischemic heart disease (etoricoxib, ibuprofen and tenoxicam) and use of lipid‐lowering drugs (diclofenac) was present.
Conclusions: The risk of AMI differed considerably for individual NSAIDs and was highest for ketorolac. An increased risk of AMI should not be considered an effect of some selective COX‐2 inhibitors only.Collaboration of databases across countries with different prescription patterns enabled to yield risk estimates for a large variety of individual NSAIDS and also of infrequently used NSAIDs.