Six studies (1,341,673 participants) were included in the review; four cohort studies and two nested case-control studies. Sample sizes ranged from 7,636 to 469,647. Four studies scored maximum points for selection of studies. All studies scored maximum points for comparability and for ascertainment of outcome in cohort studies or exposure in case-control studies.
There was an increase in all subtypes of incident stroke with use of rofecoxib (RR 1.64, 95% CI 1.15 to 2.33, four studies, Ι²=74%), and diclofenac (RR 1.27, 95% CI 1.08 to 1.48, four studies, Ι²=4%). There were no significant differences between intervention and control groups for naproxen, ibuprofen or celecoxib. There was an increased risk of ischaemic stroke with rofecoxib (RR 1.82, 95% CI 1.09 to 3.04, three studies, Ι²=91%) and diclofenac (RR 1.20, 95% CI 0.99 to 1.45, four studies, Ι²=33%). There were no significant differences reported between groups for naproxen and ibuprofen.
There was insufficient data to estimate the pooled risk rate by dose and duration and for other individual NSAIDs and non-ischaemic stroke subtypes, but details for individual studies were reported. Sensitivity analysis removing the study that reported high risk ratios for naproxen and rofecoxib was conducted but significant heterogeneity remained across studies. Removal of one study that included only men and reported current exposure differently to the other studies reduced heterogeneity, and the result for rofecoxib remained statistically significant.
There was some evidence of publication bias.