In comparative cohort studies to assess risk of myocardial infarction, stroke, gastrointestinal bleed, and renal failure in patients treated with paracetamol versus ibuprofen, results vary substantially depending on the models used to control for confounding and bias.

Large-scale propensity score matching resulted in attenuated effects and increased precision. However, substantial bias remained after large-scale propensity score matching, undermining the ability to discern or rule out an effect of exposure to paracetamol on these outcomes.

For comparisons of paracetamol versus ibuprofen, when using our methods in the Clinical Practices Research Datalink, it is only possible to distinguish true effects if those true effects are large (relative risk > 2).