Cardiovascular Risk with Non-Steroidal Anti-Inflammatory Drugs: Systematic Review of Population-Based Controlled Observational Studies

BACKGROUND: Randomised trials have highlighted the cardiovascular risks of non-steroidal anti-inflammatory drugs (NSAIDs) in high doses and sometimes atypical settings. Here, we provide estimates of the comparative risks with individual NSAIDs at typical doses in community settings. METHODS AND FINDINGS: We performed a systematic review of community-based controlled observational studies. We conducted comprehensive literature searches, extracted adjusted relative risk (RR) estimates, and pooled the estimates for major cardiovascular events associated with use of individual NSAIDs, in different doses, and in populations with low and high background risks of cardiovascular events. We also compared individual drugs in pair-wise (within study) analyses, generating ratios of RRs (RRRs). Thirty case-control studies included 184,946 cardiovascular events, and 21 cohort studies described outcomes in >2.7 million exposed individuals. Of the extensively studied drugs (ten or more studies), the highest overall risks were seen with rofecoxib, 1.45 (95% CI 1.33, 1.59), and diclofenac, 1.40 (1.27, 1.55), and the lowest with ibuprofen, 1.18 (1.11, 1.25), and naproxen, 1.09 (1.02, 1.16). In a sub-set of studies, risk was elevated with low doses of rofecoxib, 1.37 (1.20, 1.57), celecoxib, 1.26 (1.09, 1.47), and diclofenac, 1.22 (1.12, 1.33), and rose in each case with higher doses. Ibuprofen risk was seen only with higher doses. Naproxen was risk-neutral at all doses. Of the less studied drugs etoricoxib, 2.05 (1.45, 2.88), etodolac, 1.55 (1.28, 1.87), and indomethacin, 1.30 (1.19, 1.41), had the highest risks. In pair-wise comparisons, etoricoxib had a higher RR than ibuprofen, RRR = 1.68 (99% CI 1.14, 2.49), and naproxen, RRR = 1.75 (1.16, 2.64); etodolac was not significantly different from naproxen and ibuprofen. Naproxen had a significantly lower risk than ibuprofen, RRR = 0.92 (0.87, 0.99). RR estimates were constant with different background risks for cardiovascular disease a